Fix mkl-static issue for Windows. (#132401)

Fix mkl-static issue for Windows. (#130697)

Background:
We found the pytorch Windows release/2.4 performance regression: https://github.com/pytorch/pytorch/issues/130619

After some debug works, I found the pytorch Windows static mkl build options are wrong:
<img width="1049" alt="image" src="https://github.com/user-attachments/assets/38692142-bfca-4c98-8092-6e105c82bb13">
1. Thread lib is wrong.
2. Miss `openmp` lib and config.
> Debug history: https://github.com/pytorch/pytorch/issues/130619#issuecomment-2226782504 and https://github.com/pytorch/pytorch/issues/130619#issuecomment-2226418611

This PR will fix `mkl-static` build options issue.
<img width="863" alt="image" src="https://github.com/user-attachments/assets/834f6cee-7e6d-4d74-b2bc-8a270f05e429">

Reference:
<img width="482" alt="image" src="https://github.com/user-attachments/assets/8184dadb-f230-4062-a49f-51df1d7285f5">

https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-link-line-advisor.html#gs.c6izlg

Pull Request resolved: https://github.com/pytorch/pytorch/pull/130697
Approved by: https://github.com/jgong5, https://github.com/atalman

(cherry picked from commit f1456c74a09ec33d7739814e13e5745408bf8b6c)

Co-authored-by: Xu Han <xu.han@outlook.com>

.ci/docker/aotriton_version.txt

@@ -0,0 +1,5 @@
+0.6b
+manylinux_2_17
+rocm6
+04b5df8c8123f90cba3ede7e971e6fbc6040d506
+3db6ecbc915893ff967abd6e1b43bd5f54949868873be60dc802086c3863e648

.ci/docker/centos-rocm/Dockerfile

@@ -113,18 +113,18 @@ COPY triton_version.txt triton_version.txt
 RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi
 RUN rm install_triton.sh common_utils.sh triton-rocm.txt triton_version.txt
 
+# Install AOTriton (Early fail)
+COPY ./aotriton_version.txt aotriton_version.txt
+COPY ./common/common_utils.sh common_utils.sh
+COPY ./common/install_aotriton.sh install_aotriton.sh
+RUN ["/bin/bash", "-c", "./install_aotriton.sh /opt/rocm && rm -rf install_aotriton.sh aotriton_version.txt common_utils.sh"]
+ENV AOTRITON_INSTALLED_PREFIX /opt/rocm/aotriton
+
 # Install ccache/sccache (do this last, so we get priority in PATH)
 COPY ./common/install_cache.sh install_cache.sh
 ENV PATH /opt/cache/bin:$PATH
 RUN bash ./install_cache.sh && rm install_cache.sh
 
-# Install AOTriton
-COPY ci_commit_pins/aotriton.txt aotriton.txt
-COPY ./common/common_utils.sh common_utils.sh
-COPY ./common/install_aotriton.sh install_aotriton.sh
-RUN bash ./install_aotriton.sh /opt/rocm/aotriton && rm -rf install_aotriton.sh aotriton aotriton.txt common_utils.sh
-ENV AOTRITON_INSTALLED_PREFIX /opt/rocm/aotriton
-
 # Include BUILD_ENVIRONMENT environment variable in image
 ARG BUILD_ENVIRONMENT
 ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}

.ci/docker/ci_commit_pins/aotriton.txt

@@ -1 +0,0 @@
-24a3fe9cb57e5cda3c923df29743f9767194cc27

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

@@ -1 +1 @@
-01cbe5045a6898c9a925f01435c8277b2fe6afcc
+21eae954efa5bf584da70324b640288c3ee7aede

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

@@ -1 +1 @@
-b8c64f64c18d8cac598b3adb355c21e7439c21de
+aac14a3b93f11d781d1d5ebc5400b15ae8df5185

.ci/docker/common/install_aotriton.sh

@@ -4,21 +4,20 @@ set -ex
 
 source "$(dirname "${BASH_SOURCE[0]}")/common_utils.sh"
 
-AOTRITON_DIR="aotriton"
-AOTRITON_PINNED_NAME="aotriton" # No .txt extension
-AOTRITON_PINNED_COMMIT=$(get_pinned_commit ${AOTRITON_PINNED_NAME})
+TARBALL='aotriton.tar.bz2'
+# This read command alwasy returns with exit code 1
+read -d "\n" VER MANYLINUX ROCMBASE PINNED_COMMIT SHA256 < aotriton_version.txt || true
+ARCH=$(uname -m)
 AOTRITON_INSTALL_PREFIX="$1"
+AOTRITON_URL="https://github.com/ROCm/aotriton/releases/download/${VER}/aotriton-${VER}-${MANYLINUX}_${ARCH}-${ROCMBASE}.tar.bz2"
 
-git clone https://github.com/ROCm/aotriton.git "${AOTRITON_DIR}"
-cd "${AOTRITON_DIR}"
-git checkout "${AOTRITON_PINNED_COMMIT}"
-git submodule sync --recursive
-git submodule update --init --recursive --force --depth 1
-mkdir build
-cd build
-cmake .. -G Ninja -DCMAKE_INSTALL_PREFIX=./install_dir -DCMAKE_BUILD_TYPE=Release -DAOTRITON_COMPRESS_KERNEL=OFF -DAOTRITON_NO_PYTHON=ON -DAOTRITON_NO_SHARED=ON
-ninja install
-mkdir -p "${AOTRITON_INSTALL_PREFIX}"
-cp -r install_dir/* "${AOTRITON_INSTALL_PREFIX}"
-find /tmp/ -mindepth 1 -delete
-rm -rf ~/.triton
+cd "${AOTRITON_INSTALL_PREFIX}"
+# Must use -L to follow redirects
+curl -L --retry 3 -o "${TARBALL}" "${AOTRITON_URL}"
+ACTUAL_SHA256=$(sha256sum "${TARBALL}" | cut -d " " -f 1)
+if [ "${SHA256}" != "${ACTUAL_SHA256}" ]; then
+  echo -n "Error: The SHA256 of downloaded tarball is ${ACTUAL_SHA256},"
+  echo " which does not match the expected value ${SHA256}."
+  exit
+fi
+tar xf "${TARBALL}" && rm -rf "${TARBALL}"

.ci/docker/ubuntu-rocm/Dockerfile

@@ -105,18 +105,18 @@ COPY triton_version.txt triton_version.txt
 RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi
 RUN rm install_triton.sh common_utils.sh triton-rocm.txt triton_version.txt
 
+# Install AOTriton
+COPY ./aotriton_version.txt aotriton_version.txt
+COPY ./common/common_utils.sh common_utils.sh
+COPY ./common/install_aotriton.sh install_aotriton.sh
+RUN ["/bin/bash", "-c", "./install_aotriton.sh /opt/rocm && rm -rf install_aotriton.sh aotriton_version.txt common_utils.sh"]
+ENV AOTRITON_INSTALLED_PREFIX /opt/rocm/aotriton
+
 # Install ccache/sccache (do this last, so we get priority in PATH)
 COPY ./common/install_cache.sh install_cache.sh
 ENV PATH /opt/cache/bin:$PATH
 RUN bash ./install_cache.sh && rm install_cache.sh
 
-# Install AOTriton
-COPY ci_commit_pins/aotriton.txt aotriton.txt
-COPY ./common/common_utils.sh common_utils.sh
-COPY ./common/install_aotriton.sh install_aotriton.sh
-RUN bash ./install_aotriton.sh /opt/rocm/aotriton && rm -rf install_aotriton.sh aotriton aotriton.txt common_utils.sh
-ENV AOTRITON_INSTALLED_PREFIX /opt/rocm/aotriton
-
 # Include BUILD_ENVIRONMENT environment variable in image
 ARG BUILD_ENVIRONMENT
 ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}

.ci/pytorch/common_utils.sh

@@ -178,7 +178,7 @@ function install_torchrec_and_fbgemm() {
 
 function clone_pytorch_xla() {
   if [[ ! -d ./xla ]]; then
-    git clone --recursive --quiet https://github.com/pytorch/xla.git
+    git clone --recursive -b r2.4 https://github.com/pytorch/xla.git
     pushd xla
     # pin the xla hash so that we don't get broken by changes to xla
     git checkout "$(cat ../.github/ci_commit_pins/xla.txt)"

.circleci/scripts/binary_populate_env.sh

@@ -75,9 +75,8 @@ export PYTORCH_BUILD_NUMBER=1
 TRITON_VERSION=$(cat $PYTORCH_ROOT/.ci/docker/triton_version.txt)
 
 # Here PYTORCH_EXTRA_INSTALL_REQUIREMENTS is already set for the all the wheel builds hence append TRITON_CONSTRAINT
+TRITON_CONSTRAINT="platform_system == 'Linux' and platform_machine == 'x86_64' and python_version < '3.13'"
 if [[ "$PACKAGE_TYPE" =~ .*wheel.* &&  -n "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" ]]; then
-  # Only linux Python < 3.13 are supported wheels for triton
-  TRITON_CONSTRAINT="platform_system == 'Linux' and platform_machine == 'x86_64' and python_version < '3.13'"
   TRITON_REQUIREMENT="triton==${TRITON_VERSION}; ${TRITON_CONSTRAINT}"
   if [[ -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*dev.* ]]; then
       TRITON_SHORTHASH=$(cut -c1-10 $PYTORCH_ROOT/.ci/docker/ci_commit_pins/triton.txt)
@@ -87,11 +86,11 @@ if [[ "$PACKAGE_TYPE" =~ .*wheel.* &&  -n "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:
 fi
 
 # Set triton via PYTORCH_EXTRA_INSTALL_REQUIREMENTS for triton rocm package
-if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*rocm.* && $(uname) == "Linux" && "$DESIRED_PYTHON" != "3.12" ]]; then
-    TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}"
+if [[ "$PACKAGE_TYPE" =~ .*wheel.* && -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*rocm.* && $(uname) == "Linux" ]]; then
+    TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}; ${TRITON_CONSTRAINT}"
     if [[ -n "$PYTORCH_BUILD_VERSION" && "$PYTORCH_BUILD_VERSION" =~ .*dev.* ]]; then
         TRITON_SHORTHASH=$(cut -c1-10 $PYTORCH_ROOT/.ci/docker/ci_commit_pins/triton-rocm.txt)
-        TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}+${TRITON_SHORTHASH}"
+        TRITON_REQUIREMENT="pytorch-triton-rocm==${TRITON_VERSION}+${TRITON_SHORTHASH}; ${TRITON_CONSTRAINT}"
     fi
     if [[ -z "${PYTORCH_EXTRA_INSTALL_REQUIREMENTS:-}" ]]; then
         export PYTORCH_EXTRA_INSTALL_REQUIREMENTS="${TRITON_REQUIREMENT}"

.circleci/scripts/binary_upload.sh

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

.github/ci_commit_pins/audio.txt

@@ -1 +1 @@
-1980f8af5bcd0bb2ce51965cf79d8d4c25dad8a0
+b829e936f7cc61b48149f5f957a451a38bf2a178

.github/ci_commit_pins/torchbench.txt

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

.github/ci_commit_pins/xla.txt

@@ -1 +1 @@
-6f0b61e5d782913a0fc7743812f2a8e522189111
+r2.4

.github/scripts/amd/package_triton_wheel.sh

@@ -93,6 +93,8 @@ done
 
 # Copy Include Files
 cp -r $ROCM_HOME/include/hip $TRITON_ROCM_DIR/include
+cp -r $ROCM_HOME/include/roctracer $TRITON_ROCM_DIR/include
+cp -r $ROCM_HOME/include/hsa $TRITON_ROCM_DIR/include
 
 # Copy linker
 mkdir -p $TRITON_ROCM_DIR/llvm/bin

.github/scripts/filter_test_configs.py

@@ -38,9 +38,9 @@ SUPPORTED_PERIODICAL_MODES: Dict[str, Callable[[Optional[str]], bool]] = {
 }
 
 # The link to the published list of disabled jobs
-DISABLED_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/disabled-jobs.json"
+DISABLED_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/disabled-jobs.json?versionId=tIl0Qo224T_NDVw0dtG4hU1cZJM97inV"
 # and unstable jobs
-UNSTABLE_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/unstable-jobs.json"
+UNSTABLE_JOBS_URL = "https://ossci-metrics.s3.amazonaws.com/unstable-jobs.json?versionId=GPyRZRsOo26Gfk_WjAoNNxEMGXkIxIes"
 
 # Some constants used to handle disabled and unstable jobs
 JOB_NAME_SEP = "/"

.github/scripts/generate_binary_build_matrix.py

@@ -347,10 +347,6 @@ def generate_wheels_matrix(
     for python_version in python_versions:
         for arch_version in arches:
             gpu_arch_type = arch_type(arch_version)
-            # Disable py3.12 builds for ROCm because of triton dependency
-            # on llnl-hatchet, which doesn't have py3.12 wheels available
-            if gpu_arch_type == "rocm" and python_version == "3.12":
-                continue
             gpu_arch_version = (
                 ""
                 if arch_version == "cpu"
@@ -390,6 +386,29 @@ def generate_wheels_matrix(
                         ),
                     }
                 )
+                # Special build building to use on Colab. PyThon 3.10 for 12.1 CUDA
+                if (
+                    arch_version != "cuda-aarch64"
+                    and python_version == "3.10"
+                    and arch_version == "12.1"
+                ):
+                    ret.append(
+                        {
+                            "python_version": python_version,
+                            "gpu_arch_type": gpu_arch_type,
+                            "gpu_arch_version": gpu_arch_version,
+                            "desired_cuda": translate_desired_cuda(
+                                gpu_arch_type, gpu_arch_version
+                            ),
+                            "devtoolset": "",
+                            "container_image": WHEEL_CONTAINER_IMAGES[arch_version],
+                            "package_type": package_type,
+                            "pytorch_extra_install_requirements": "",
+                            "build_name": f"{package_type}-py{python_version}-{gpu_arch_type}{gpu_arch_version}-full".replace(  # noqa: B950
+                                ".", "_"
+                            ),
+                        }
+                    )
             else:
                 ret.append(
                     {

.github/templates/common.yml.j2

@@ -8,7 +8,7 @@
 # NOTE: If testing pytorch/builder changes you can change this variable to change what pytorch/builder reference
 #       the binary builds will check out
 {%- set builder_repo = "pytorch/builder" -%}
-{%- set builder_branch = "main" -%}
+{%- set builder_branch = "release/2.4" -%}
 
 {%- macro concurrency(build_environment) -%}
 concurrency:

.github/templates/linux_binary_build_workflow.yml.j2

@@ -113,8 +113,8 @@ jobs:
         with:
           name: !{{ config["build_name"] }}
           path: "${{ runner.temp }}/artifacts/"
-      !{{ common.checkout(deep_clone=False, directory="pytorch") }}
-      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch) }}
+      !{{ common.checkout(deep_clone=False, directory="pytorch", checkout_pr_head=False) }}
+      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch, checkout_pr_head=False) }}
       - name: ROCm set GPU_FLAG
         run: |
           echo "GPU_FLAG=--device=/dev/mem --device=/dev/kfd --device=/dev/dri --group-add video --group-add daemon" >> "${GITHUB_ENV}"

.github/templates/macos_binary_build_workflow.yml.j2

@@ -81,8 +81,8 @@ jobs:
           elif [ -d "/Applications/Xcode_13.3.1.app" ]; then
             echo "DEVELOPER_DIR=/Applications/Xcode_13.3.1.app/Contents/Developer" >> "${GITHUB_ENV}"
           fi
-      !{{ common.checkout(deep_clone=False, directory="pytorch") }}
-      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch) }}
+      !{{ common.checkout(deep_clone=False, directory="pytorch", checkout_pr_head=False) }}
+      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch, checkout_pr_head=False) }}
       - name: Install sccache (only for non-forked PRs, and pushes to trunk)
         uses: nick-fields/retry@v2.8.2
         if: ${{ github.event_name == 'push' || github.event.pull_request.head.repo.full_name == github.repository }}

.github/templates/windows_binary_build_workflow.yml.j2

@@ -65,8 +65,8 @@ jobs:
     steps:
       !{{ common.setup_ec2_windows() }}
       !{{ set_runner_specific_vars() }}
-      !{{ common.checkout(deep_clone=False, directory="pytorch") }}
-      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch) }}
+      !{{ common.checkout(deep_clone=False, directory="pytorch", checkout_pr_head=False) }}
+      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch, checkout_pr_head=False) }}
       - name: Populate binary env
         shell: bash
         run: |
@@ -105,8 +105,8 @@ jobs:
         with:
           name: !{{ config["build_name"] }}
           path: "${{ env.PYTORCH_FINAL_PACKAGE_DIR }}"
-      !{{ common.checkout(deep_clone=False, directory="pytorch") }}
-      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch) }}
+      !{{ common.checkout(deep_clone=False, directory="pytorch", checkout_pr_head=False) }}
+      !{{ common.checkout(deep_clone=False, directory="builder", repository=common.builder_repo, branch=common.builder_branch, checkout_pr_head=False) }}
       - name: Populate binary env
         shell: bash
         run: |

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

@@ -37,7 +37,7 @@ jobs:
       keep-going: ${{ steps.filter.outputs.keep-going }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false
@@ -59,25 +59,25 @@ jobs:
     runs-on: ${{ matrix.runner }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image-name }}
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -141,5 +141,5 @@ jobs:
         if: always()
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

.github/workflows/_android-full-build-test.yml

@@ -37,7 +37,7 @@ jobs:
       keep-going: ${{ steps.filter.outputs.keep-going }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false
@@ -59,25 +59,25 @@ jobs:
     runs-on: ${{ matrix.runner }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image-name }}
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -186,5 +186,5 @@ jobs:
         if: always()
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

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

@@ -42,7 +42,7 @@ jobs:
       reenabled-issues: ${{ steps.filter.outputs.reenabled-issues }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false
@@ -64,25 +64,25 @@ jobs:
     runs-on: ${{ matrix.runner }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image-name }}
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -92,7 +92,7 @@ jobs:
         run: echo "IN_ARC_RUNNER=$([ -f /.inarc ] && echo true || echo false)" >> "$GITHUB_OUTPUT"
 
       - name: Install nvidia driver, nvidia-docker runtime, set GPU_FLAG
-        uses: pytorch/test-infra/.github/actions/setup-nvidia@main
+        uses: pytorch/test-infra/.github/actions/setup-nvidia@release/2.4
         if: ${{ inputs.cuda-version != 'cpu' && steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
 
       - name: Output disk space left
@@ -201,5 +201,5 @@ jobs:
           file-suffix: bazel-${{ github.job }}_${{ steps.get-job-id.outputs.job-id }}
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

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

@@ -145,13 +145,13 @@ jobs:
 
       - name: "[FB EMPLOYEES] Enable SSH (Click me for login details)"
         if: inputs.build_environment != 'linux-s390x-binary-manywheel'
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         continue-on-error: true
         with:
           github-secret: ${{ secrets.github-token }}
 
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           no-sudo: ${{ inputs.build_environment == 'linux-aarch64-binary-manywheel' || inputs.build_environment == 'linux-s390x-binary-manywheel' }}
 
@@ -181,7 +181,6 @@ jobs:
       - name: Checkout PyTorch to pytorch dir
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -195,7 +194,7 @@ jobs:
       - name: Checkout pytorch/builder to builder dir
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -221,7 +220,7 @@ jobs:
 
       - name: Pull Docker image
         if: ${{ steps.filter.outputs.is-test-matrix-empty == 'False' && inputs.build_environment != 'linux-s390x-binary-manywheel' }}
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ inputs.DOCKER_IMAGE }}
 
@@ -278,7 +277,7 @@ jobs:
 
       - name: Teardown Linux
         if: always() && inputs.build_environment != 'linux-s390x-binary-manywheel'
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
 
       - name: Chown workspace
         if: always() && inputs.build_environment != 'linux-s390x-binary-manywheel'

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

@@ -128,14 +128,14 @@ jobs:
 
       - name: "[FB EMPLOYEES] Enable SSH (Click me for login details)"
         if: inputs.build_environment != 'linux-s390x-binary-manywheel'
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         continue-on-error: true
         with:
           github-secret: ${{ secrets.github-token }}
 
         # Setup the environment
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           no-sudo: ${{ inputs.build_environment == 'linux-aarch64-binary-manywheel' || inputs.build_environment == 'linux-s390x-binary-manywheel' }}
 
@@ -158,7 +158,6 @@ jobs:
       - name: Checkout PyTorch to pytorch dir
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
 
@@ -171,7 +170,7 @@ jobs:
       - name: Checkout pytorch/builder to builder dir
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -202,12 +201,12 @@ jobs:
           path: "${{ runner.temp }}/artifacts/"
 
       - name: Install nvidia driver, nvidia-docker runtime, set GPU_FLAG
-        uses: pytorch/test-infra/.github/actions/setup-nvidia@main
+        uses: pytorch/test-infra/.github/actions/setup-nvidia@release/2.4
         if: ${{ inputs.GPU_ARCH_TYPE == 'cuda' && steps.filter.outputs.is-test-matrix-empty == 'False' }}
 
       - name: Pull Docker image
         if: ${{ steps.filter.outputs.is-test-matrix-empty == 'False' && inputs.build_environment != 'linux-s390x-binary-manywheel' }}
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ inputs.DOCKER_IMAGE }}
 
@@ -217,7 +216,7 @@ jobs:
 
       - name: Teardown Linux
         if: always() && inputs.build_environment != 'linux-s390x-binary-manywheel'
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
 
       - name: Chown workspace
         if: always() && inputs.build_environment != 'linux-s390x-binary-manywheel'

.github/workflows/_binary-upload.yml

@@ -95,7 +95,7 @@ jobs:
       SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           no-sudo: true
 

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

@@ -23,7 +23,7 @@ jobs:
       keep-going: ${{ steps.filter.outputs.keep-going }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false
@@ -44,7 +44,7 @@ jobs:
     runs-on: ${{ matrix.runner }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Set up JDK 8
         uses: actions/setup-java@v3
@@ -53,7 +53,7 @@ jobs:
           distribution: 'temurin'
 
       - name: Setup miniconda
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: 3.8
           environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}

.github/workflows/_docs.yml

@@ -80,7 +80,7 @@ jobs:
     name: build-docs-${{ matrix.docs_type }}-${{ inputs.push }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
           instructions: |
@@ -91,7 +91,7 @@ jobs:
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
@@ -106,12 +106,12 @@ jobs:
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image }}
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -218,5 +218,5 @@ jobs:
           s3-prefix: pytorch/pytorch/${{ github.event.pull_request.number }}/functorchdocs
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

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

@@ -46,7 +46,7 @@ jobs:
       keep-going: ${{ steps.filter.outputs.keep-going }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false
@@ -80,7 +80,7 @@ jobs:
     steps:
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Populate CI build options
         shell: bash
@@ -102,7 +102,7 @@ jobs:
             brew install libtool
 
       - name: Setup miniconda for iOS
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: "3.9"
           environment-file: .github/requirements/conda-env-iOS.txt

.github/workflows/_linux-build-label.yml

@@ -81,7 +81,7 @@ jobs:
       test-matrix: ${{ steps.linux-build.outputs.test-matrix }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
@@ -90,7 +90,7 @@ jobs:
       # checkout because when we run this action we don't *have* a local
       # checkout. In other cases you should prefer a local checkout.
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Linux Build
         id: linux-build

.github/workflows/_linux-build-rg.yml

@@ -86,7 +86,7 @@ jobs:
       # checkout because when we run this action we don't *have* a local
       # checkout. In other cases you should prefer a local checkout.
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Linux Build
         id: linux-build

.github/workflows/_linux-build.yml

@@ -90,7 +90,7 @@ jobs:
       test-matrix: ${{ steps.filter.outputs.test-matrix }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
@@ -99,7 +99,7 @@ jobs:
       # checkout because when we run this action we don't *have* a local
       # checkout. In other cases you should prefer a local checkout.
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
@@ -114,7 +114,7 @@ jobs:
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image-name }}
 
@@ -128,7 +128,7 @@ jobs:
           echo "docker pull ghcr.io/pytorch/ci-image:${tag/:/-}"
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -238,5 +238,5 @@ jobs:
           s3-bucket: ${{ inputs.s3-bucket }}
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

.github/workflows/_linux-test-label.yml

@@ -67,7 +67,7 @@ jobs:
     timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Linux Test
         id: linux-test

.github/workflows/_linux-test-rg.yml

@@ -68,7 +68,7 @@ jobs:
     timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Linux Test
         id: linux-test

.github/workflows/_linux-test.yml

@@ -67,7 +67,7 @@ jobs:
     timeout-minutes: ${{ matrix.mem_leak_check == 'mem_leak_check' && 600 || inputs.timeout-minutes }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         if: ${{ !contains(matrix.runner, 'gcp.a100') }}
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
@@ -76,7 +76,7 @@ jobs:
               docker exec -it $(docker container ps --format '{{.ID}}') bash
 
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
@@ -91,7 +91,7 @@ jobs:
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image }}
 
@@ -105,7 +105,7 @@ jobs:
           echo "docker pull ghcr.io/pytorch/ci-image:${tag/:/-}"
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
@@ -116,7 +116,7 @@ jobs:
 
       - name: Install nvidia driver, nvidia-docker runtime, set GPU_FLAG
         id: install-nvidia-driver
-        uses: pytorch/test-infra/.github/actions/setup-nvidia@main
+        uses: pytorch/test-infra/.github/actions/setup-nvidia@release/2.4
         if: ${{ contains(inputs.build-environment, 'cuda') && !contains(matrix.config, 'nogpu') && steps.check_arc_runner.outputs.IN_ARC_RUNNER == 'false' }}
 
       - name: Lock NVIDIA A100 40GB Frequency
@@ -333,7 +333,7 @@ jobs:
           path: ./**/core.[1-9]*
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()
 
       # NB: We are currently having an intermittent GPU-related issue on G5 runners with

.github/workflows/_mac-build.yml

@@ -71,11 +71,11 @@ jobs:
       test-matrix: ${{ steps.filter.outputs.test-matrix }}
     steps:
       - name: Clean up disk space before running MacOS workflow
-        uses: pytorch/test-infra/.github/actions/check-disk-space@main
+        uses: pytorch/test-infra/.github/actions/check-disk-space@release/2.4
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Set xcode version
         env:
@@ -87,7 +87,7 @@ jobs:
 
       - name: Setup miniconda
         if: inputs.environment-file == ''
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: ${{ inputs.python-version }}
           environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}
@@ -97,7 +97,7 @@ jobs:
       # environment even though the arch is x86-64
       - name: Setup miniconda using the provided environment file
         if: inputs.environment-file != ''
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: ${{ inputs.python-version }}
           environment-file: ${{ inputs.environment-file }}
@@ -207,4 +207,4 @@ jobs:
       - name: Clean up disk space
         if: always()
         continue-on-error: true
-        uses: pytorch/test-infra/.github/actions/check-disk-space@main
+        uses: pytorch/test-infra/.github/actions/check-disk-space@release/2.4

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

@@ -40,7 +40,7 @@ jobs:
       reenabled-issues: ${{ steps.filter.outputs.reenabled-issues }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
 
@@ -81,7 +81,7 @@ jobs:
           use-gha: true
 
       - name: Setup miniconda
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: ${{ inputs.python-version }}
           environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}
@@ -159,4 +159,4 @@ jobs:
       - name: Clean up disk space
         if: always()
         continue-on-error: true
-        uses: pytorch/test-infra/.github/actions/check-disk-space@main
+        uses: pytorch/test-infra/.github/actions/check-disk-space@release/2.4

.github/workflows/_mac-test.yml

@@ -74,11 +74,11 @@ jobs:
           done
 
       - name: Clean up disk space before running MacOS workflow
-        uses: pytorch/test-infra/.github/actions/check-disk-space@main
+        uses: pytorch/test-infra/.github/actions/check-disk-space@release/2.4
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Download build artifacts
         uses: ./.github/actions/download-build-artifacts
@@ -93,7 +93,7 @@ jobs:
           use-gha: true
 
       - name: Setup miniconda
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: ${{ inputs.python-version }}
           environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}
@@ -216,4 +216,4 @@ jobs:
       - name: Clean up disk space
         if: always()
         continue-on-error: true
-        uses: pytorch/test-infra/.github/actions/check-disk-space@main
+        uses: pytorch/test-infra/.github/actions/check-disk-space@release/2.4

.github/workflows/_rocm-test.yml

@@ -58,7 +58,7 @@ jobs:
     steps:
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           no-sudo: true
 
@@ -80,12 +80,12 @@ jobs:
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image }}
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 

.github/workflows/_run_android_tests.yml

@@ -23,7 +23,7 @@ jobs:
       keep-going: ${{ steps.filter.outputs.keep-going }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false
@@ -54,10 +54,10 @@ jobs:
       SUPPORT_ABI: '${{ matrix.support_abi }}'
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup miniconda
-        uses: pytorch/test-infra/.github/actions/setup-miniconda@main
+        uses: pytorch/test-infra/.github/actions/setup-miniconda@release/2.4
         with:
           python-version: 3.8
           environment-file: .github/requirements/conda-env-${{ runner.os }}-${{ runner.arch }}.txt

.github/workflows/_runner-determinator.yml

@@ -32,7 +32,7 @@ jobs:
       USERNAME: ${{ inputs.user_name }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: true

.github/workflows/_win-build.yml

@@ -60,10 +60,10 @@ jobs:
           git config --global core.fsmonitor false
 
       - name: Clean up leftover processes on non-ephemeral Windows runner
-        uses: pytorch/test-infra/.github/actions/cleanup-runner@main
+        uses: pytorch/test-infra/.github/actions/cleanup-runner@release/2.4
 
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
           instructions: |
@@ -78,7 +78,7 @@ jobs:
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           no-sudo: true
 

.github/workflows/_win-test.yml

@@ -54,10 +54,10 @@ jobs:
           git config --global core.fsmonitor false
 
       - name: Clean up leftover processes on non-ephemeral Windows runner
-        uses: pytorch/test-infra/.github/actions/cleanup-runner@main
+        uses: pytorch/test-infra/.github/actions/cleanup-runner@release/2.4
 
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
           instructions: |
@@ -73,7 +73,7 @@ jobs:
 
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           no-sudo: true
 

.github/workflows/_xpu-test.yml

@@ -54,7 +54,7 @@ jobs:
     steps:
       # [see note: pytorch repo ref]
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup XPU
         uses: ./.github/actions/setup-xpu
@@ -72,12 +72,12 @@ jobs:
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ inputs.docker-image }}
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 

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

@@ -3,7 +3,7 @@ name: Build Triton wheels
 on:
   push:
     branches:
-      - main
+      - release/2.4
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
       # Release candidate tags look like: v1.11.0-rc1
@@ -47,12 +47,12 @@ jobs:
       BUILD_DEVICE: ${{ matrix.device }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
 
@@ -60,7 +60,7 @@ jobs:
         uses: ./.github/actions/setup-linux
 
       - name: Pull Docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ env.DOCKER_IMAGE }}
 
@@ -124,7 +124,7 @@ jobs:
           path: ${{ runner.temp }}/artifacts/*
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()
 
   upload-wheel:
@@ -209,12 +209,12 @@ jobs:
       PY_VERS: ${{ matrix.py_vers }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
 
@@ -222,7 +222,7 @@ jobs:
         uses: ./.github/actions/setup-linux
 
       - name: Pull Docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ env.DOCKER_IMAGE }}
 
@@ -257,7 +257,7 @@ jobs:
           path: ${{ runner.temp }}/artifacts/*
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()
 
   upload-conda:

.github/workflows/check-labels.yml

@@ -31,7 +31,7 @@ jobs:
     runs-on: linux.20_04.4x
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
           fetch-depth: 1

.github/workflows/close-nonexistent-disable-issues.yml

@@ -11,7 +11,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Run close_nonexistent_disable_issues.py
         env:

.github/workflows/create_release.yml

@@ -5,6 +5,11 @@ on:
     branches:
       - main
       - release/*
+    tags:
+      # Final Release tags look like: v1.11.0
+      - v[0-9]+.[0-9]+.[0-9]+
+      # Release candidate tags look like: v1.11.0-rc1
+      - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   release:
     types: [published]
   pull_request:
@@ -18,6 +23,8 @@ jobs:
     # https://github.com/softprops/action-gh-release?tab=readme-ov-file#permissions
     permissions:
       contents: write
+    outputs:
+      pt_release_name: ${{ steps.release_name.outputs.pt_release_name }}
     steps:
       - uses: malfet/checkout@silent-checkout
         with:
@@ -49,11 +56,44 @@ jobs:
             # Create archive
             tar -czf "$PT_RELEASE_FILE" "$PT_RELEASE_NAME"
             echo "Created source archive $PT_RELEASE_FILE with content: $(ls -a "$PT_RELEASE_NAME")"
-      - name: Upload source distribution
+      - name: Upload source distribution for release
         if: ${{ github.event_name == 'release' }}
         uses: softprops/action-gh-release@v1
         with:
           files: ${{env.PT_RELEASE_FILE}}
+      - name: Upload source distribution to GHA artifacts for release tags
+        if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') && contains(github.ref, 'rc') }}
+        uses: actions/upload-artifact@v2
+        with:
+          name: ${{ env.PT_RELEASE_FILE }}
+          path: ${{ env.PT_RELEASE_FILE }}
+      - name: Set output
+        id: release_name
+        run: echo "::set-output name=pt_release_name::${{ env.PT_RELEASE_NAME }}.tar.gz"
+
+  upload_source_code_to_s3:
+    if: ${{ github.repository == 'pytorch/pytorch' && github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') && contains(github.ref, 'rc') }}
+    runs-on: linux.2xlarge
+    environment: sourcecode-upload
+    name: Upload source code to S3 for release tags
+    permissions:
+      id-token: write
+    needs: release
+    steps:
+      - uses: actions/download-artifact@v2
+        with:
+          name: ${{ needs.release.outputs.pt_release_name }}
+      - name: Configure AWS credentials(PyTorch account)
+        uses: aws-actions/configure-aws-credentials@v3
+        with:
+          role-to-assume: arn:aws:iam::749337293305:role/gha_pytorch_source_code_upload_role
+          aws-region: us-east-1
+      - uses: seemethere/upload-artifact-s3@v5
+        with:
+          s3-bucket: pytorch
+          s3-prefix: source_code/test
+          if-no-files-found: warn
+          path: ${{ needs.release.outputs.pt_release_name }}
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name }}

.github/workflows/docker-builds.yml

@@ -78,21 +78,21 @@ jobs:
       # [see note: pytorch repo ref]
       # deep clone (fetch-depth 0) required for git merge-base
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - name: Setup Linux
         uses: ./.github/actions/setup-linux
 
       - name: Build docker image
         id: build-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: ${{ matrix.docker-image-name }}
           always-rebuild: true
           push: true
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.build-docker-image.outputs.docker-image }}
 
@@ -124,5 +124,5 @@ jobs:
         if: always()
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

.github/workflows/docker-release.yml

@@ -41,7 +41,7 @@ jobs:
       matrix: ${{ steps.generate-matrix.outputs.matrix }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: true
@@ -69,7 +69,7 @@ jobs:
       CUDNN_VERSION: ${{ matrix.cudnn_version }}
     steps:
       - name: Setup SSH (Click me for login details)
-        uses: pytorch/test-infra/.github/actions/setup-ssh@main
+        uses: pytorch/test-infra/.github/actions/setup-ssh@release/2.4
         with:
           github-secret: ${{ secrets.GITHUB_TOKEN }}
       # [see note: pytorch repo ref]
@@ -147,12 +147,12 @@ jobs:
           fi
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()
 
   validate:
     needs: build
-    uses: pytorch/builder/.github/workflows/validate-docker-images.yml@main
+    uses: pytorch/builder/.github/workflows/validate-docker-images.yml@release/2.4
     with:
       channel: nightly
       ref: main

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

@@ -48,7 +48,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.8"
       runs_on: linux.arm64.m7g.4xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
@@ -69,7 +69,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
@@ -91,7 +91,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-aarch64
     secrets:
@@ -111,7 +111,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.8"
       runs_on: linux.arm64.m7g.4xlarge
@@ -135,7 +135,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda-aarch64
@@ -156,7 +156,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.9"
       runs_on: linux.arm64.m7g.4xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
@@ -177,7 +177,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
@@ -199,7 +199,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-aarch64
     secrets:
@@ -219,7 +219,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.9"
       runs_on: linux.arm64.m7g.4xlarge
@@ -243,7 +243,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cuda-aarch64
@@ -264,7 +264,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.10"
       runs_on: linux.arm64.m7g.4xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
@@ -285,7 +285,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
@@ -307,7 +307,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-aarch64
     secrets:
@@ -327,7 +327,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.10"
       runs_on: linux.arm64.m7g.4xlarge
@@ -351,7 +351,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cuda-aarch64
@@ -372,7 +372,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.11"
       runs_on: linux.arm64.m7g.4xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
@@ -393,7 +393,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
@@ -415,7 +415,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-aarch64
     secrets:
@@ -435,7 +435,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.11"
       runs_on: linux.arm64.m7g.4xlarge
@@ -459,7 +459,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cuda-aarch64
@@ -480,7 +480,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.12"
       runs_on: linux.arm64.m7g.4xlarge
       ALPINE_IMAGE: "arm64v8/alpine"
@@ -501,7 +501,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-aarch64
       build_environment: linux-aarch64-binary-manywheel
@@ -523,7 +523,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cpu-aarch64-2.4
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-aarch64
     secrets:
@@ -543,7 +543,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.12"
       runs_on: linux.arm64.m7g.4xlarge
@@ -567,7 +567,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cu124
       GPU_ARCH_TYPE: cuda-aarch64
-      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.4-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cuda-aarch64

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

@@ -48,7 +48,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cpu
       build_environment: linux-binary-conda
@@ -66,7 +66,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cpu
       build_environment: linux-binary-conda
@@ -87,7 +87,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cpu
     secrets:
@@ -108,7 +108,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.8"
       runs_on: linux.24xlarge
       build_name: conda-py3_8-cuda11_8
@@ -128,7 +128,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cuda11_8
       build_environment: linux-binary-conda
@@ -150,7 +150,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cuda11_8
     secrets:
@@ -171,7 +171,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.8"
       runs_on: linux.24xlarge
       build_name: conda-py3_8-cuda12_1
@@ -191,7 +191,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cuda12_1
       build_environment: linux-binary-conda
@@ -213,7 +213,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cuda12_1
     secrets:
@@ -234,7 +234,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.8"
       runs_on: linux.24xlarge
       build_name: conda-py3_8-cuda12_4
@@ -254,7 +254,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cuda12_4
       build_environment: linux-binary-conda
@@ -276,7 +276,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cuda12_4
     secrets:
@@ -296,7 +296,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cpu
       build_environment: linux-binary-conda
@@ -314,7 +314,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cpu
       build_environment: linux-binary-conda
@@ -335,7 +335,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cpu
     secrets:
@@ -356,7 +356,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.9"
       runs_on: linux.24xlarge
       build_name: conda-py3_9-cuda11_8
@@ -376,7 +376,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cuda11_8
       build_environment: linux-binary-conda
@@ -398,7 +398,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cuda11_8
     secrets:
@@ -419,7 +419,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.9"
       runs_on: linux.24xlarge
       build_name: conda-py3_9-cuda12_1
@@ -439,7 +439,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cuda12_1
       build_environment: linux-binary-conda
@@ -461,7 +461,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cuda12_1
     secrets:
@@ -482,7 +482,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.9"
       runs_on: linux.24xlarge
       build_name: conda-py3_9-cuda12_4
@@ -502,7 +502,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cuda12_4
       build_environment: linux-binary-conda
@@ -524,7 +524,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cuda12_4
     secrets:
@@ -544,7 +544,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cpu
       build_environment: linux-binary-conda
@@ -562,7 +562,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cpu
       build_environment: linux-binary-conda
@@ -583,7 +583,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cpu
     secrets:
@@ -604,7 +604,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.10"
       runs_on: linux.24xlarge
       build_name: conda-py3_10-cuda11_8
@@ -624,7 +624,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cuda11_8
       build_environment: linux-binary-conda
@@ -646,7 +646,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cuda11_8
     secrets:
@@ -667,7 +667,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.10"
       runs_on: linux.24xlarge
       build_name: conda-py3_10-cuda12_1
@@ -687,7 +687,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cuda12_1
       build_environment: linux-binary-conda
@@ -709,7 +709,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cuda12_1
     secrets:
@@ -730,7 +730,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.10"
       runs_on: linux.24xlarge
       build_name: conda-py3_10-cuda12_4
@@ -750,7 +750,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cuda12_4
       build_environment: linux-binary-conda
@@ -772,7 +772,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cuda12_4
     secrets:
@@ -792,7 +792,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cpu
       build_environment: linux-binary-conda
@@ -810,7 +810,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cpu
       build_environment: linux-binary-conda
@@ -831,7 +831,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cpu
     secrets:
@@ -852,7 +852,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.11"
       runs_on: linux.24xlarge
       build_name: conda-py3_11-cuda11_8
@@ -872,7 +872,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cuda11_8
       build_environment: linux-binary-conda
@@ -894,7 +894,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cuda11_8
     secrets:
@@ -915,7 +915,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.11"
       runs_on: linux.24xlarge
       build_name: conda-py3_11-cuda12_1
@@ -935,7 +935,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cuda12_1
       build_environment: linux-binary-conda
@@ -957,7 +957,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cuda12_1
     secrets:
@@ -978,7 +978,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.11"
       runs_on: linux.24xlarge
       build_name: conda-py3_11-cuda12_4
@@ -998,7 +998,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cuda12_4
       build_environment: linux-binary-conda
@@ -1020,7 +1020,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cuda12_4
     secrets:
@@ -1040,7 +1040,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cpu
       build_environment: linux-binary-conda
@@ -1058,7 +1058,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cpu
       build_environment: linux-binary-conda
@@ -1079,7 +1079,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cpu
     secrets:
@@ -1100,7 +1100,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.12"
       runs_on: linux.24xlarge
       build_name: conda-py3_12-cuda11_8
@@ -1120,7 +1120,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cuda11_8
       build_environment: linux-binary-conda
@@ -1142,7 +1142,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cuda11_8
     secrets:
@@ -1163,7 +1163,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.12"
       runs_on: linux.24xlarge
       build_name: conda-py3_12-cuda12_1
@@ -1183,7 +1183,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cuda12_1
       build_environment: linux-binary-conda
@@ -1205,7 +1205,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cuda12_1
     secrets:
@@ -1226,7 +1226,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.12"
       runs_on: linux.24xlarge
       build_name: conda-py3_12-cuda12_4
@@ -1246,7 +1246,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cuda12_4
       build_environment: linux-binary-conda
@@ -1268,7 +1268,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/conda-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cuda12_4
     secrets:

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

@@ -43,7 +43,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cpu-shared-with-deps-cxx11-abi
@@ -62,7 +62,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cpu-shared-with-deps-cxx11-abi

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

@@ -48,7 +48,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cpu-shared-with-deps-cxx11-abi
@@ -67,7 +67,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cpu-shared-with-deps-cxx11-abi
@@ -89,7 +89,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cpu-shared-with-deps-cxx11-abi
@@ -111,7 +111,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda11.8-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda11_8-shared-with-deps-cxx11-abi
@@ -131,7 +131,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda11.8-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda11_8-shared-with-deps-cxx11-abi
@@ -154,7 +154,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda11.8-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda11_8-shared-with-deps-cxx11-abi
@@ -176,7 +176,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda12_1-shared-with-deps-cxx11-abi
@@ -196,7 +196,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda12_1-shared-with-deps-cxx11-abi
@@ -219,7 +219,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda12_1-shared-with-deps-cxx11-abi
@@ -241,7 +241,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.4-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda12_4-shared-with-deps-cxx11-abi
@@ -261,7 +261,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.4-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda12_4-shared-with-deps-cxx11-abi
@@ -284,7 +284,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cuda12.4-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cuda12_4-shared-with-deps-cxx11-abi
@@ -306,7 +306,7 @@ jobs:
       DESIRED_CUDA: rocm6.0
       GPU_ARCH_VERSION: 6.0
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.0-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.0-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-rocm6_0-shared-with-deps-cxx11-abi
@@ -328,7 +328,7 @@ jobs:
       GPU_ARCH_VERSION: 6.0
       GPU_ARCH_TYPE: rocm
       SKIP_ALL_TESTS: 1
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.0-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.0-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
     steps:
@@ -342,7 +342,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -354,7 +353,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -370,7 +369,7 @@ jobs:
       - name: Pull Docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
         with:
-          docker-image: pytorch/libtorch-cxx11-builder:rocm6.0-main
+          docker-image: pytorch/libtorch-cxx11-builder:rocm6.0-2.4
       - name: Test Pytorch binary
         uses: ./pytorch/.github/actions/test-pytorch-binary
       - name: Teardown ROCm
@@ -390,7 +389,7 @@ jobs:
       DESIRED_CUDA: rocm6.0
       GPU_ARCH_VERSION: 6.0
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.0-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.0-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-rocm6_0-shared-with-deps-cxx11-abi
@@ -412,7 +411,7 @@ jobs:
       DESIRED_CUDA: rocm6.1
       GPU_ARCH_VERSION: 6.1
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.1-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-rocm6_1-shared-with-deps-cxx11-abi
@@ -434,7 +433,7 @@ jobs:
       GPU_ARCH_VERSION: 6.1
       GPU_ARCH_TYPE: rocm
       SKIP_ALL_TESTS: 1
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.1-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
     steps:
@@ -448,7 +447,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -460,7 +458,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -476,7 +474,7 @@ jobs:
       - name: Pull Docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
         with:
-          docker-image: pytorch/libtorch-cxx11-builder:rocm6.1-main
+          docker-image: pytorch/libtorch-cxx11-builder:rocm6.1-2.4
       - name: Test Pytorch binary
         uses: ./pytorch/.github/actions/test-pytorch-binary
       - name: Teardown ROCm
@@ -496,7 +494,7 @@ jobs:
       DESIRED_CUDA: rocm6.1
       GPU_ARCH_VERSION: 6.1
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.1-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:rocm6.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-rocm6_1-shared-with-deps-cxx11-abi

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

@@ -43,7 +43,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cpu-shared-with-deps-pre-cxx11
@@ -62,7 +62,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cpu-shared-with-deps-pre-cxx11

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

@@ -48,7 +48,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cpu-shared-with-deps-pre-cxx11
@@ -67,7 +67,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cpu-shared-with-deps-pre-cxx11
@@ -89,7 +89,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cpu-shared-with-deps-pre-cxx11
@@ -111,7 +111,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda11_8-shared-with-deps-pre-cxx11
@@ -131,7 +131,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda11_8-shared-with-deps-pre-cxx11
@@ -154,7 +154,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda11_8-shared-with-deps-pre-cxx11
@@ -176,7 +176,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda12_1-shared-with-deps-pre-cxx11
@@ -196,7 +196,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda12_1-shared-with-deps-pre-cxx11
@@ -219,7 +219,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda12_1-shared-with-deps-pre-cxx11
@@ -241,7 +241,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda12_4-shared-with-deps-pre-cxx11
@@ -261,7 +261,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda12_4-shared-with-deps-pre-cxx11
@@ -284,7 +284,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-cuda12_4-shared-with-deps-pre-cxx11
@@ -306,7 +306,7 @@ jobs:
       DESIRED_CUDA: rocm6.0
       GPU_ARCH_VERSION: 6.0
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.0-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.0-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-rocm6_0-shared-with-deps-pre-cxx11
@@ -328,7 +328,7 @@ jobs:
       GPU_ARCH_VERSION: 6.0
       GPU_ARCH_TYPE: rocm
       SKIP_ALL_TESTS: 1
-      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.0-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.0-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
     steps:
@@ -342,7 +342,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -354,7 +353,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -370,7 +369,7 @@ jobs:
       - name: Pull Docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
         with:
-          docker-image: pytorch/manylinux-builder:rocm6.0-main
+          docker-image: pytorch/manylinux-builder:rocm6.0-2.4
       - name: Test Pytorch binary
         uses: ./pytorch/.github/actions/test-pytorch-binary
       - name: Teardown ROCm
@@ -390,7 +389,7 @@ jobs:
       DESIRED_CUDA: rocm6.0
       GPU_ARCH_VERSION: 6.0
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.0-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.0-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-rocm6_0-shared-with-deps-pre-cxx11
@@ -412,7 +411,7 @@ jobs:
       DESIRED_CUDA: rocm6.1
       GPU_ARCH_VERSION: 6.1
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-rocm6_1-shared-with-deps-pre-cxx11
@@ -434,7 +433,7 @@ jobs:
       GPU_ARCH_VERSION: 6.1
       GPU_ARCH_TYPE: rocm
       SKIP_ALL_TESTS: 1
-      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
     steps:
@@ -448,7 +447,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -460,7 +458,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -476,7 +474,7 @@ jobs:
       - name: Pull Docker image
         uses: pytorch/test-infra/.github/actions/pull-docker-image@main
         with:
-          docker-image: pytorch/manylinux-builder:rocm6.1-main
+          docker-image: pytorch/manylinux-builder:rocm6.1-2.4
       - name: Test Pytorch binary
         uses: ./pytorch/.github/actions/test-pytorch-binary
       - name: Teardown ROCm
@@ -496,7 +494,7 @@ jobs:
       DESIRED_CUDA: rocm6.1
       GPU_ARCH_VERSION: 6.1
       GPU_ARCH_TYPE: rocm
-      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:rocm6.1-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: pre-cxx11
       build_name: libtorch-rocm6_1-shared-with-deps-pre-cxx11

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

@@ -44,7 +44,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda11_8
       build_environment: linux-binary-manywheel
@@ -64,7 +64,7 @@ jobs:
       DESIRED_CUDA: cu118
       GPU_ARCH_VERSION: 11.8
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda11.8-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda11_8
       build_environment: linux-binary-manywheel
@@ -84,7 +84,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda12_1
       build_environment: linux-binary-manywheel
@@ -104,7 +104,7 @@ jobs:
       DESIRED_CUDA: cu121
       GPU_ARCH_VERSION: 12.1
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.1-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda12_1
       build_environment: linux-binary-manywheel
@@ -124,7 +124,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda12_4
       build_environment: linux-binary-manywheel
@@ -144,7 +144,7 @@ jobs:
       DESIRED_CUDA: cu124
       GPU_ARCH_VERSION: 12.4
       GPU_ARCH_TYPE: cuda
-      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cuda12.4-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cuda12_4
       build_environment: linux-binary-manywheel

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

@@ -48,7 +48,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.8"
       runs_on: linux.s390x
       ALPINE_IMAGE: "docker.io/s390x/alpine"
@@ -69,7 +69,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
@@ -91,7 +91,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-s390x
     secrets:
@@ -111,7 +111,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.9"
       runs_on: linux.s390x
       ALPINE_IMAGE: "docker.io/s390x/alpine"
@@ -132,7 +132,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
@@ -154,7 +154,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-s390x
     secrets:
@@ -174,7 +174,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.10"
       runs_on: linux.s390x
       ALPINE_IMAGE: "docker.io/s390x/alpine"
@@ -195,7 +195,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
@@ -217,7 +217,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-s390x
     secrets:
@@ -237,7 +237,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.11"
       runs_on: linux.s390x
       ALPINE_IMAGE: "docker.io/s390x/alpine"
@@ -258,7 +258,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
@@ -280,7 +280,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-s390x
     secrets:
@@ -300,7 +300,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.12"
       runs_on: linux.s390x
       ALPINE_IMAGE: "docker.io/s390x/alpine"
@@ -321,7 +321,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-s390x
       build_environment: linux-s390x-binary-manywheel
@@ -343,7 +343,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu-s390x
-      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-main
+      DOCKER_IMAGE: pytorch/manylinuxs390x-builder:cpu-s390x-2.4
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-s390x
     secrets:

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

@@ -77,7 +77,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -89,7 +88,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -141,7 +140,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.8"
       build_name: conda-py3_8-cpu
       use_s3: False
@@ -195,7 +194,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -207,7 +205,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -259,7 +257,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.9"
       build_name: conda-py3_9-cpu
       use_s3: False
@@ -313,7 +311,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -325,7 +322,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -377,7 +374,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.10"
       build_name: conda-py3_10-cpu
       use_s3: False
@@ -431,7 +428,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -443,7 +439,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -495,7 +491,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.11"
       build_name: conda-py3_11-cpu
       use_s3: False
@@ -549,7 +545,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -561,7 +556,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -613,7 +608,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/conda-builder:cpu-main
+      DOCKER_IMAGE: pytorch/conda-builder:cpu-2.4
       DESIRED_PYTHON: "3.12"
       build_name: conda-py3_12-cpu
       use_s3: False

.github/workflows/generated-macos-arm64-binary-libtorch-cxx11-abi-nightly.yml

@@ -81,7 +81,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -93,7 +92,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -145,7 +144,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-main
+      DOCKER_IMAGE: pytorch/libtorch-cxx11-builder:cpu-2.4
       LIBTORCH_VARIANT: shared-with-deps
       DESIRED_DEVTOOLSET: cxx11-abi
       build_name: libtorch-cpu-shared-with-deps-cxx11-abi

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

@@ -78,7 +78,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -90,7 +89,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -142,7 +141,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       DESIRED_PYTHON: "3.8"
       build_name: wheel-py3_8-cpu
       use_s3: False
@@ -197,7 +196,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -209,7 +207,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -261,7 +259,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       DESIRED_PYTHON: "3.9"
       build_name: wheel-py3_9-cpu
       use_s3: False
@@ -316,7 +314,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -328,7 +325,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -380,7 +377,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       DESIRED_PYTHON: "3.10"
       build_name: wheel-py3_10-cpu
       use_s3: False
@@ -435,7 +432,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -447,7 +443,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -499,7 +495,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       DESIRED_PYTHON: "3.11"
       build_name: wheel-py3_11-cpu
       use_s3: False
@@ -554,7 +550,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -566,7 +561,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -618,7 +613,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu
       GPU_ARCH_TYPE: cpu
-      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-main
+      DOCKER_IMAGE: pytorch/manylinux-builder:cpu-2.4
       DESIRED_PYTHON: "3.12"
       build_name: wheel-py3_12-cpu
       use_s3: False

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

@@ -93,7 +93,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -105,7 +104,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -210,7 +209,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -222,7 +220,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -336,7 +334,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -348,7 +345,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -454,7 +451,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -466,7 +462,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -581,7 +577,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -593,7 +588,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -699,7 +694,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -711,7 +705,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -826,7 +820,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -838,7 +831,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -944,7 +937,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -956,7 +948,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1070,7 +1062,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1082,7 +1073,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1187,7 +1178,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1199,7 +1189,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1313,7 +1303,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1325,7 +1314,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1431,7 +1420,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1443,7 +1431,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1558,7 +1546,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1570,7 +1557,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1676,7 +1663,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1688,7 +1674,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1803,7 +1789,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1815,7 +1800,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1921,7 +1906,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1933,7 +1917,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2047,7 +2031,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2059,7 +2042,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2164,7 +2147,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2176,7 +2158,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2290,7 +2272,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2302,7 +2283,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2408,7 +2389,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2420,7 +2400,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2535,7 +2515,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2547,7 +2526,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2653,7 +2632,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2665,7 +2643,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2780,7 +2758,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2792,7 +2769,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2898,7 +2875,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2910,7 +2886,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3024,7 +3000,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3036,7 +3011,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3141,7 +3116,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3153,7 +3127,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3267,7 +3241,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3279,7 +3252,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3385,7 +3358,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3397,7 +3369,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3512,7 +3484,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3524,7 +3495,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3630,7 +3601,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3642,7 +3612,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3757,7 +3727,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3769,7 +3738,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3875,7 +3844,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3887,7 +3855,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4001,7 +3969,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4013,7 +3980,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4118,7 +4085,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4130,7 +4096,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4244,7 +4210,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4256,7 +4221,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4362,7 +4327,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4374,7 +4338,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4489,7 +4453,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4501,7 +4464,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4607,7 +4570,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4619,7 +4581,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4734,7 +4696,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4746,7 +4707,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4852,7 +4813,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4864,7 +4824,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder

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

@@ -90,7 +90,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -102,7 +101,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -211,7 +210,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -223,7 +221,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder

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

@@ -97,7 +97,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -109,7 +108,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -218,7 +217,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -230,7 +228,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -352,7 +350,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -364,7 +361,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -474,7 +471,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -486,7 +482,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -609,7 +605,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -621,7 +616,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -731,7 +726,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -743,7 +737,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -866,7 +860,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -878,7 +871,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -988,7 +981,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1000,7 +992,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder

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

@@ -90,7 +90,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -102,7 +101,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -211,7 +210,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -223,7 +221,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder

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

@@ -97,7 +97,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -109,7 +108,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -218,7 +217,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -230,7 +228,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -352,7 +350,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -364,7 +361,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -474,7 +471,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -486,7 +482,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -609,7 +605,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -621,7 +616,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -731,7 +726,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -743,7 +737,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -866,7 +860,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -878,7 +871,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -988,7 +981,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1000,7 +992,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder

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

@@ -94,7 +94,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -106,7 +105,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -211,7 +210,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -223,7 +221,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -338,7 +336,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -350,7 +347,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -456,7 +453,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -468,7 +464,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -584,7 +580,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -596,7 +591,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -702,7 +697,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -714,7 +708,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -830,7 +824,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -842,7 +835,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -948,7 +941,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -960,7 +952,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1075,7 +1067,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1087,7 +1078,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1192,7 +1183,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1204,7 +1194,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1319,7 +1309,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1331,7 +1320,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1437,7 +1426,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1449,7 +1437,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1565,7 +1553,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1577,7 +1564,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1683,7 +1670,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1695,7 +1681,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1811,7 +1797,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1823,7 +1808,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1929,7 +1914,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -1941,7 +1925,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2056,7 +2040,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2068,7 +2051,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2173,7 +2156,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2185,7 +2167,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2300,7 +2282,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2312,7 +2293,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2418,7 +2399,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2430,7 +2410,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2546,7 +2526,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2558,7 +2537,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2664,7 +2643,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2676,7 +2654,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2792,7 +2770,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2804,7 +2781,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2910,7 +2887,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -2922,7 +2898,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3037,7 +3013,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3049,7 +3024,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3154,7 +3129,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3166,7 +3140,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3281,7 +3255,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3293,7 +3266,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3399,7 +3372,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3411,7 +3383,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3527,7 +3499,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3539,7 +3510,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3645,7 +3616,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3657,7 +3627,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3773,7 +3743,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3785,7 +3754,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -3891,7 +3860,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -3903,7 +3871,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4018,7 +3986,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4030,7 +3997,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4135,7 +4102,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4147,7 +4113,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4262,7 +4228,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4274,7 +4239,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4380,7 +4345,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4392,7 +4356,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4508,7 +4472,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4520,7 +4483,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4626,7 +4589,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4638,7 +4600,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4754,7 +4716,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4766,7 +4727,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -4872,7 +4833,6 @@ jobs:
       - name: Checkout PyTorch
         uses: malfet/checkout@silent-checkout
         with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           submodules: recursive
           path: pytorch
           quiet-checkout: true
@@ -4884,7 +4844,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder

.github/workflows/lint-bc.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Run BC Lint Action
-        uses: pytorch/test-infra/.github/actions/bc-lint@main
+        uses: pytorch/test-infra/.github/actions/bc-lint@release/2.4
         with:
           repo: ${{ github.event.pull_request.head.repo.full_name }}
           base_sha: ${{ github.event.pull_request.base.sha }}

.github/workflows/lint.yml

@@ -16,7 +16,7 @@ permissions: read-all
 # When any other step fails, it's job will be retried once by retryBot.
 jobs:
   lintrunner-clang:
-    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@release/2.4
     with:
       timeout: 120
       runner: linux.2xlarge
@@ -32,7 +32,7 @@ jobs:
         .github/scripts/lintrunner.sh
 
   lintrunner-noclang:
-    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@release/2.4
     with:
       timeout: 120
       runner: linux.2xlarge
@@ -47,7 +47,7 @@ jobs:
         .github/scripts/lintrunner.sh
 
   quick-checks:
-    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@release/2.4
     with:
       runner: linux.2xlarge
       docker-image: pytorch-linux-focal-linter
@@ -88,7 +88,7 @@ jobs:
     if: github.event_name == 'pull_request' && !contains(github.event.pull_request.labels.*.name, 'skip-pr-sanity-checks')
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
           fetch-depth: -1
@@ -101,7 +101,7 @@ jobs:
           bash .github/scripts/pr-sanity-check.sh
 
   workflow-checks:
-    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@release/2.4
     with:
       runner: linux.2xlarge
       docker-image: pytorch-linux-focal-linter
@@ -112,6 +112,7 @@ jobs:
         # The generic Linux job chooses to use base env, not the one setup by the image
         CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
         conda activate "${CONDA_ENV}"
+        export RELEASE_VERSION_TAG="2.4"
 
         # Regenerate workflows
         .github/scripts/generate_ci_workflows.py
@@ -137,7 +138,7 @@ jobs:
         exit $RC
 
   toc:
-    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@release/2.4
     with:
       runner: linux.2xlarge
       docker-image: pytorch-linux-focal-linter
@@ -175,7 +176,7 @@ jobs:
   test-tools:
     name: Test tools
     if: ${{ github.repository == 'pytorch/pytorch' }}
-    uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
+    uses: pytorch/test-infra/.github/workflows/linux_job.yml@release/2.4
     with:
       runner: linux.2xlarge
       docker-image: pytorch-linux-focal-linter
@@ -196,7 +197,7 @@ jobs:
     runs-on: linux.20_04.4x
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
           fetch-depth: 1
@@ -226,7 +227,7 @@ jobs:
       # [see note: pytorch repo ref]
       # deep clone (fetch-depth 0) required, to allow us to use git log
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
           fetch-depth: 1

.github/workflows/llm_td_retrieval.yml

@@ -116,5 +116,5 @@ jobs:
           AWS_REGION: ""
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()

.github/workflows/nightly-rockset-uploads.yml

@@ -21,7 +21,7 @@ jobs:
     environment: upload-stats
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false

.github/workflows/nightly.yml

@@ -41,7 +41,7 @@ jobs:
     environment: update-commit-hash
     steps:
       - name: update-vision-commit-hash
-        uses: pytorch/test-infra/.github/actions/update-commit-hash@main
+        uses: pytorch/test-infra/.github/actions/update-commit-hash@release/2.4
         if: ${{ github.event_name == 'schedule' }}
         with:
           repo-name: vision
@@ -56,7 +56,7 @@ jobs:
     environment: update-commit-hash
     steps:
       - name: update-audio-commit-hash
-        uses: pytorch/test-infra/.github/actions/update-commit-hash@main
+        uses: pytorch/test-infra/.github/actions/update-commit-hash@release/2.4
         if: ${{ github.event_name == 'schedule' }}
         with:
           repo-name: audio
@@ -71,7 +71,7 @@ jobs:
     environment: update-commit-hash
     steps:
       - name: update-executorch-commit-hash
-        uses: pytorch/test-infra/.github/actions/update-commit-hash@main
+        uses: pytorch/test-infra/.github/actions/update-commit-hash@release/2.4
         if: ${{ github.event_name == 'schedule' }}
         with:
           repo-name: executorch

.github/workflows/target-determination-indexer.yml

@@ -24,7 +24,7 @@ jobs:
 
       - name: Calculate docker image
         id: calculate-docker-image
-        uses: pytorch/test-infra/.github/actions/calculate-docker-image@main
+        uses: pytorch/test-infra/.github/actions/calculate-docker-image@release/2.4
         with:
           docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9
           working-directory: pytorch
@@ -39,13 +39,13 @@ jobs:
           echo "docker pull ghcr.io/pytorch/ci-image:${tag/:/-}"
 
       - name: Pull docker image
-        uses: pytorch/test-infra/.github/actions/pull-docker-image@main
+        uses: pytorch/test-infra/.github/actions/pull-docker-image@release/2.4
         with:
           docker-image: ${{ steps.calculate-docker-image.outputs.docker-image }}
 
       - name: Install nvidia driver, nvidia-docker runtime, set GPU_FLAG
         id: install-nvidia-driver
-        uses: pytorch/test-infra/.github/actions/setup-nvidia@main
+        uses: pytorch/test-infra/.github/actions/setup-nvidia@release/2.4
 
       - name: Clone CodeLlama
         uses: actions/checkout@v3
@@ -136,7 +136,7 @@ jobs:
             "s3://target-determinator-assets/indexes/latest/${ZIP_NAME}"
 
       - name: Teardown Linux
-        uses: pytorch/test-infra/.github/actions/teardown-linux@main
+        uses: pytorch/test-infra/.github/actions/teardown-linux@release/2.4
         if: always()
 
 concurrency:

.github/workflows/target_determination.yml

@@ -14,7 +14,7 @@ jobs:
       # checkout because when we run this action we don't *have* a local
       # checkout. In other cases you should prefer a local checkout.
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
 

.github/workflows/update-viablestrict.yml

@@ -16,7 +16,7 @@ jobs:
     environment: ${{ (github.event_name == 'schedule') && 'mergebot' || '' }}
     steps:
       - name: Update viable/strict
-        uses: pytorch/test-infra/.github/actions/update-viablestrict@main
+        uses: pytorch/test-infra/.github/actions/update-viablestrict@release/2.4
         with:
           repository: pytorch/pytorch
           stable-branch: viable/strict

.github/workflows/update_pytorch_labels.yml

@@ -17,7 +17,7 @@ jobs:
       contents: read
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false

.github/workflows/upload-alerts.yml

@@ -44,7 +44,7 @@ jobs:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
           AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-        uses: pytorch/test-infra/.github/actions/upload-alerts@main
+        uses: pytorch/test-infra/.github/actions/upload-alerts@release/2.4
         with:
           alerts: '${{ steps.alert_creation_step.outputs.script-output }}'
           organization: "pytorch"

.github/workflows/upload-test-stats.yml

@@ -39,7 +39,7 @@ jobs:
         run: echo "${TRIGGERING_WORKFLOW}"
 
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
 
       - uses: actions/setup-python@v4
         with:

.github/workflows/upload-torch-dynamo-perf-stats.yml

@@ -29,7 +29,7 @@ jobs:
     name: Upload dynamo performance stats for ${{ github.event.workflow_run.id }}, attempt ${{ github.event.workflow_run.run_attempt }}
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           submodules: false
           fetch-depth: 1

.github/workflows/upload_test_stats_intermediate.yml

@@ -17,7 +17,7 @@ jobs:
     environment: upload-stats
     steps:
       - name: Checkout PyTorch
-        uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@release/2.4
         with:
           fetch-depth: 1
           submodules: false

.github/workflows/weekly.yml

@@ -21,7 +21,7 @@ jobs:
           fetch-depth: 0
       - name: update-xla-commit-hash
         continue-on-error: true
-        uses: pytorch/test-infra/.github/actions/update-commit-hash@main
+        uses: pytorch/test-infra/.github/actions/update-commit-hash@release/2.4
         with:
           repo-name: xla
           branch: master
@@ -30,7 +30,7 @@ jobs:
           updatebot-token: ${{ secrets.UPDATEBOT_TOKEN }}
           pytorchbot-token: ${{ secrets.GH_PYTORCHBOT_TOKEN }}
       - name: update-triton-commit-hash
-        uses: pytorch/test-infra/.github/actions/update-commit-hash@main
+        uses: pytorch/test-infra/.github/actions/update-commit-hash@release/2.4
         with:
           repo-owner: openai
           repo-name: triton

CMakeLists.txt

@@ -865,12 +865,13 @@ cmake_dependent_option(
 # Suspect users building from source will need this
 add_definitions(-DFLASHATTENTION_DISABLE_ALIBI)
 
-# CAVEAT: Again, do not check USE_ROCM here Flash Attention2 will error while
-# building for sm52 while Mem Eff Attention won't
+# CAVEAT: Again, Flash Attention2 will error while building for sm52 while Mem
+# Eff Attention won't
 cmake_dependent_option(
   USE_MEM_EFF_ATTENTION
   "Enable memory-efficient attention for scaled dot product attention.\
-  Will be disabled if not supported by the platform" ON "USE_CUDA" OFF)
+  Will be disabled if not supported by the platform" ON
+  "USE_CUDA OR USE_ROCM" OFF)
 
 if(DEBUG_CUDA)
   string(APPEND CMAKE_CUDA_FLAGS_DEBUG " -lineinfo")

aten/src/ATen/TensorIndexing.h

@@ -197,7 +197,7 @@ TORCH_API std::ostream& operator<<(
     const std::vector<TensorIndex>& tensor_indices);
 
 namespace impl {
-static inline Tensor applySlice(
+inline Tensor applySlice(
     const Tensor& self,
     int64_t dim,
     c10::SymInt start,
@@ -227,7 +227,7 @@ static inline Tensor applySlice(
       dim, std::move(start), std::move(stop), std::move(step));
 }
 
-static inline Tensor applySelect(
+inline Tensor applySelect(
     const Tensor& self,
     int64_t dim,
     SymInt index,
@@ -266,9 +266,7 @@ static inline Tensor applySelect(
   return self.select_symint(dim, std::move(index));
 }
 
-static inline Tensor boolToIndexingTensorCPUOrCUDA(
-    const Tensor& self,
-    bool value) {
+inline Tensor boolToIndexingTensorCPUOrCUDA(const Tensor& self, bool value) {
   // booleans add a dimension of size 1. true indexes this dimension as if 0:,
   // false as empty.
   if (value) {
@@ -278,7 +276,7 @@ static inline Tensor boolToIndexingTensorCPUOrCUDA(
   }
 }
 
-static inline Tensor boolToIndexingTensorNonNativeDeviceType(
+inline Tensor boolToIndexingTensorNonNativeDeviceType(
     const Tensor& self,
     bool value) {
   // booleans add a dimension of size 1. true indexes this dimension as if 0:,
@@ -290,7 +288,7 @@ static inline Tensor boolToIndexingTensorNonNativeDeviceType(
   }
 }
 
-static inline Tensor boolToIndexingTensor(
+inline Tensor boolToIndexingTensor(
     const Tensor& self,
     bool value,
     const at::Device& self_device) {
@@ -301,13 +299,13 @@ static inline Tensor boolToIndexingTensor(
   }
 }
 
-static inline Tensor scalarToTensorNonNativeDeviceType(
+inline Tensor scalarToTensorNonNativeDeviceType(
     const Scalar& v,
     const TensorOptions& options) {
   return at::scalar_tensor(v, options);
 }
 
-static inline void recordTensorIndex(
+inline void recordTensorIndex(
     const Tensor& tensor,
     std::vector<Tensor>& outIndices,
     int64_t* dim_ptr) {
@@ -317,7 +315,7 @@ static inline void recordTensorIndex(
   (*dim_ptr)++;
 };
 
-static inline c10::List<::std::optional<Tensor>> typeConvertIndices(
+inline c10::List<::std::optional<Tensor>> typeConvertIndices(
     const Tensor& /*self*/,
     std::vector<Tensor>&& indices) {
   c10::List<::std::optional<Tensor>> converted_inds;
@@ -338,7 +336,7 @@ static inline c10::List<::std::optional<Tensor>> typeConvertIndices(
 // construct a `std::vector` container to be consumed by the C++
 // `count_specified_dimensions` function, which adds 100s of nanoseconds
 // overhead and is undesirable.
-static inline int64_t count_specified_dimensions(
+inline int64_t count_specified_dimensions(
     const ArrayRef<TensorIndex>& indices) {
   // Count the number of indexed dimensions (everything but ellipsis and None)
   int64_t count = 0;
@@ -372,7 +370,7 @@ static inline int64_t count_specified_dimensions(
 //
 // The rest of the functions are in `at::indexing::impl` namespace, signifying
 // that they shouldn't be used from Python indexing implementation.
-static inline Tensor scalarToTensor(
+inline Tensor scalarToTensor(
     const Scalar& v,
     const TensorOptions& options,
     const at::Device& self_device) {
@@ -387,7 +385,7 @@ static inline Tensor scalarToTensor(
 // To match numpy semantics:
 // As a special case for backwards compatibility,
 // strip away unit dimensions from the left of 'src'
-static inline SymIntArrayRef slicePrefix1sSize(const SymIntArrayRef& sizes) {
+inline SymIntArrayRef slicePrefix1sSize(const SymIntArrayRef& sizes) {
   size_t first_non1_src = sizes.size();
   for (const auto i : c10::irange(sizes.size())) {
     // Unbacked SymInt has different behavior, but this is sound because
@@ -402,7 +400,7 @@ static inline SymIntArrayRef slicePrefix1sSize(const SymIntArrayRef& sizes) {
   return sizes.slice(first_non1_src);
 }
 
-static inline void copy_to(const Tensor& dst, const Tensor& src) {
+inline void copy_to(const Tensor& dst, const Tensor& src) {
   if (dst.sym_sizes().equals(src.sym_sizes())) {
     // A shortcut to avoid generating hard-coded constant sizes during tracing.
     // This is not a perfect solution: when src & dst have different shapes,
@@ -421,7 +419,7 @@ static inline void copy_to(const Tensor& dst, const Tensor& src) {
 
 // See NOTE [ Setting `disable_slice_optimization` when calling C++ tensor
 // indexing functions from Python ]
-static inline Tensor handleDimInMultiDimIndexing(
+inline Tensor handleDimInMultiDimIndexing(
     const Tensor& prev_dim_result,
     const Tensor& original_tensor,
     const TensorIndex& index,
@@ -509,7 +507,7 @@ static inline Tensor handleDimInMultiDimIndexing(
 namespace impl {
 // This mirrors `applySlicing` in
 // torch/csrc/autograd/python_variable_indexing.cpp
-static inline Tensor applySlicing(
+inline Tensor applySlicing(
     const Tensor& self,
     const ArrayRef<TensorIndex>& indices,
     std::vector<Tensor>& outIndices,
@@ -550,13 +548,13 @@ static inline Tensor applySlicing(
 }
 } // namespace impl
 
-static inline Tensor dispatch_index(
+inline Tensor dispatch_index(
     const Tensor& self,
     std::vector<Tensor>&& indices) {
   return self.index(impl::typeConvertIndices(self, std::move(indices)));
 }
 
-static inline Tensor dispatch_index_put_(
+inline Tensor dispatch_index_put_(
     Tensor& self,
     std::vector<Tensor>&& indices,
     const Tensor& value) {
@@ -598,7 +596,7 @@ static inline Tensor dispatch_index_put_(
 // torch/csrc/autograd/python_variable_indexing.cpp See NOTE [ Setting
 // `disable_slice_optimization` when calling C++ tensor indexing functions from
 // Python ]
-static inline Tensor get_item(
+inline Tensor get_item(
     const Tensor& self,
     const ArrayRef<TensorIndex>& indices,
     bool disable_slice_optimization = false) {
@@ -664,7 +662,7 @@ static inline Tensor get_item(
 // torch/csrc/autograd/python_variable_indexing.cpp for "the assigned value is a
 // Tensor" case See NOTE [ Setting `disable_slice_optimization` when calling C++
 // tensor indexing functions from Python ]
-static inline void set_item(
+inline void set_item(
     const Tensor& self,
     const ArrayRef<TensorIndex>& indices,
     const Tensor& value,

aten/src/ATen/core/Generator.h

@@ -150,7 +150,7 @@ Generator make_generator(Args&&... args) {
  * the backend generator type (CPU/CUDAGeneratorImpl etc.)
  */
 template <typename T>
-static inline T * check_generator(std::optional<Generator> gen) {
+inline T * check_generator(std::optional<Generator> gen) {
   TORCH_CHECK(gen.has_value(), "Expected Generator but received nullopt");
   TORCH_CHECK(gen->defined(), "Generator with undefined implementation is not allowed");
   TORCH_CHECK(T::device_type() == gen->device().type(), "Expected a '", T::device_type(), "' device type for generator but found '", gen->device().type(), "'");
@@ -164,7 +164,7 @@ static inline T * check_generator(std::optional<Generator> gen) {
  * the backend generator type (CPU/CUDAGeneratorImpl etc.)
  */
 template <typename T>
-static inline T* get_generator_or_default(const std::optional<Generator>& gen, const Generator& default_gen) {
+inline T* get_generator_or_default(const std::optional<Generator>& gen, const Generator& default_gen) {
   return gen.has_value() && gen->defined() ? check_generator<T>(gen) : check_generator<T>(default_gen);
 }
 
@@ -177,7 +177,7 @@ namespace detail {
  * - The new state tensor must be a torch.ByteTensor
  * - Data of the new state tensor must be contiguous
  */
-static inline void check_rng_state(const c10::TensorImpl& new_state) {
+inline void check_rng_state(const c10::TensorImpl& new_state) {
   TORCH_CHECK_TYPE(
     new_state.layout() == kStrided && new_state.device().type() == kCPU && new_state.dtype() == kByte,
     "RNG state must be a torch.ByteTensor"

aten/src/ATen/core/MetaFallbackKernel.cpp

@@ -17,7 +17,7 @@ static void metaFallback(
       "while using an operator with PT2 compilation APIs (torch.compile/torch.export); "
       "in order to use this operator with those APIs you'll need to add a fake impl. "
       "Please see the following for next steps:  "
-      "https://pytorch.org/docs/main/notes/custom_operators.html");
+      "https://pytorch.org/tutorials/advanced/custom_ops_landing_page.html");
 }
 
 TORCH_LIBRARY_IMPL(_, Meta, m) {

aten/src/ATen/core/TensorBase.h

@@ -953,7 +953,7 @@ TensorBase make_tensor_base(Args&&... args) {
 
 } // namespace detail
 
-static inline DispatchKey legacyExtractDispatchKey(const TensorBase& t) {
+inline DispatchKey legacyExtractDispatchKey(const TensorBase& t) {
   return legacyExtractDispatchKey(t.key_set());
 }
 

aten/src/ATen/core/dispatch/DispatchKeyExtractor.h

@@ -21,7 +21,7 @@ namespace impl {
 // on TLS.
 //
 // NB: If there is no valid dispatch key, this will return Undefined
-static inline DispatchKeySet computeDispatchKeySet(
+inline DispatchKeySet computeDispatchKeySet(
     DispatchKeySet ks,
     // The key mask lets us eliminate (by zero entries) keys which should not
     // be considered for dispatch.  There are two cases when we use this:

aten/src/ATen/core/stack.h

@@ -66,51 +66,51 @@ class Operation {
 
 // treat the last N elements of the stack as a list, looking up
 // element i
-static inline IValue& peek(Stack& stack, size_t i, size_t N) {
+inline IValue& peek(Stack& stack, size_t i, size_t N) {
   // NOLINTNEXTLINE(cppcoreguidelines-narrowing-conversions)
   return *(stack.end() - N + i);
 }
-static inline IValue& peek(Stack* stack, size_t i, size_t N) {
+inline IValue& peek(Stack* stack, size_t i, size_t N) {
   return peek(*stack, i, N);
 }
-static inline const IValue& peek(const Stack& stack, size_t i, size_t N) {
+inline const IValue& peek(const Stack& stack, size_t i, size_t N) {
   // NOLINTNEXTLINE(cppcoreguidelines-narrowing-conversions)
   return *(stack.end() - N + i);
 }
-static inline const IValue& peek(const Stack* stack, size_t i, size_t N) {
+inline const IValue& peek(const Stack* stack, size_t i, size_t N) {
   return peek(*stack, i, N);
 }
 // treat the last N elements of the stack as a list, looking up the
 // slice starting at index i and having length len
-static inline at::ArrayRef<IValue> peekSlice(
+inline at::ArrayRef<IValue> peekSlice(
     const Stack& stack,
     size_t i,
     size_t len,
     size_t N) {
   return at::ArrayRef<IValue>(stack).slice(stack.size() - N + i, len);
 }
-static inline at::ArrayRef<IValue> last(const Stack& stack, size_t N) {
+inline at::ArrayRef<IValue> last(const Stack& stack, size_t N) {
   return peekSlice(stack, 0, N, N);
 }
-static inline at::ArrayRef<IValue> last(const Stack* stack, size_t N) {
+inline at::ArrayRef<IValue> last(const Stack* stack, size_t N) {
   return last(*stack, N);
 }
-static inline void drop(Stack& stack, size_t n) {
+inline void drop(Stack& stack, size_t n) {
   // NOLINTNEXTLINE(cppcoreguidelines-narrowing-conversions)
   stack.erase(stack.end() - n, stack.end());
 }
-static inline void drop(Stack* stack, size_t n) {
+inline void drop(Stack* stack, size_t n) {
   drop(*stack, n);
 }
-static inline IValue pop(Stack& stack) {
+inline IValue pop(Stack& stack) {
   auto r = std::move(stack.back());
   stack.pop_back();
   return r;
 }
-static inline IValue pop(Stack* stack) {
+inline IValue pop(Stack* stack) {
   return pop(*stack);
 }
-static inline std::vector<IValue> pop(Stack& stack, size_t n) {
+inline std::vector<IValue> pop(Stack& stack, size_t n) {
   std::vector<IValue> result;
   result.reserve(n);
   for (const auto i : c10::irange(n)) {
@@ -127,7 +127,7 @@ static inline std::vector<IValue> pop(Stack& stack, size_t n) {
 // b = pop(stack).toTensor();
 // a = pop(stack).toInt();
 template <typename... Types>
-static inline void pop(Stack& stack, Types&... args) {
+inline void pop(Stack& stack, Types&... args) {
   size_t i = 0;
   constexpr size_t N = sizeof...(args);
   (void)std::initializer_list<int>{
@@ -135,15 +135,15 @@ static inline void pop(Stack& stack, Types&... args) {
   drop(stack, N);
 }
 template <typename... Types>
-static inline void pop(Stack* stack, Types&... args) {
+inline void pop(Stack* stack, Types&... args) {
   pop(*stack, args...);
 }
 template <typename Type>
-static inline void push_one(Stack& stack, Type&& arg) {
+inline void push_one(Stack& stack, Type&& arg) {
   stack.emplace_back(std::forward<Type>(arg));
 }
 
-static inline void push_one(Stack& stack, c10::TensorOptions options) {
+inline void push_one(Stack& stack, c10::TensorOptions options) {
   stack.emplace_back(c10::typeMetaToScalarType(options.dtype()));
   stack.emplace_back(options.layout());
   stack.emplace_back(options.device());
@@ -151,15 +151,15 @@ static inline void push_one(Stack& stack, c10::TensorOptions options) {
 }
 
 template <typename... Types>
-static inline void push(Stack& stack, Types&&... args) {
+inline void push(Stack& stack, Types&&... args) {
   (void)std::initializer_list<int>{(push_one(stack, std::forward<Types>(args)), 0)...};
 }
 template <typename... Types>
-static inline void push(Stack* stack, Types&&... args) {
+inline void push(Stack* stack, Types&&... args) {
   return push(*stack, std::forward<Types>(args)...);
 }
 template <class T>
-static inline void push_list_elements(Stack& stack, const c10::List<T>& elements) {
+inline void push_list_elements(Stack& stack, const c10::List<T>& elements) {
   for (T elem : elements) {
     stack.push_back(std::move(elem));
   }

aten/src/ATen/cuda/CUDAGeneratorImpl.cpp

@@ -152,9 +152,6 @@ void CUDAGeneratorState::register_graph(cuda::CUDAGraph* graph) {
  * Unregisters a CUDA graph from the RNG state.
  */
 void CUDAGeneratorState::unregister_graph(cuda::CUDAGraph* graph) {
-  // Ensures that the RNG state is not currently being captured.
-  at::cuda::assertNotCapturing(
-      "Cannot unregister the state during capturing stage.");
   // Verify the graph was previously registered.
   TORCH_CHECK(
       registered_graphs_.find(graph) != registered_graphs_.end(),

aten/src/ATen/cuda/tunable/GemmCommon.h

@@ -84,11 +84,23 @@ struct GemmParams : OpParams {
     return c10::str(transa, transb, "_", m, "_", n, "_", k);
   }
 
+  size_t GetSizeA() const {
+    return sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m);
+  }
+
+  size_t GetSizeB() const {
+    return sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k);
+  }
+
+  size_t GetSizeC() const {
+    return sizeof(T) * ldc * n;
+  }
+
   size_t GetSize(bool duplicate_inputs) const {
-    size_t size = sizeof(T) * ldc * n;
+    size_t size = GetSizeC();
     if (duplicate_inputs) {
-      size += sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m);
-      size += sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k);
+      size += GetSizeA();
+      size += GetSizeB();
     }
     return size;
   }
@@ -98,13 +110,13 @@ struct GemmParams : OpParams {
     *copy = *this;
     c10::DeviceIndex device = 0;
     AT_CUDA_CHECK(c10::cuda::GetDevice(&device));
-    size_t c_size = ldc * n * sizeof(T);
+    size_t c_size = GetSizeC();
     copy->c = static_cast<T*>(c10::cuda::CUDACachingAllocator::raw_alloc(c_size));
     AT_CUDA_CHECK(c10::cuda::CUDACachingAllocator::memcpyAsync(
         copy->c, device, c, device, c_size, getCurrentCUDAStream(device), true));
     if (duplicate_inputs) {
-      size_t a_size = sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m);
-      size_t b_size = sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k);
+      size_t a_size = GetSizeA();
+      size_t b_size = GetSizeB();
       copy->a = static_cast<const T*>(c10::cuda::CUDACachingAllocator::raw_alloc(a_size));
       copy->b = static_cast<const T*>(c10::cuda::CUDACachingAllocator::raw_alloc(b_size));
       copy->duplicate_inputs_ = true;
@@ -153,11 +165,23 @@ struct GemmStridedBatchedParams : OpParams {
     return c10::str(transa, transb, "_", m, "_", n, "_", k, "_B_", batch);
   }
 
+  size_t GetSizeA() const {
+    return sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m) * batch;
+  }
+
+  size_t GetSizeB() const {
+    return sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k) * batch;
+  }
+
+  size_t GetSizeC() const {
+    return sizeof(T) * ldc * n * batch;
+  }
+
   size_t GetSize(bool duplicate_inputs) const {
-    size_t size = sizeof(T) * stride_c * batch;
+    size_t size = GetSizeC();
     if (duplicate_inputs) {
-      size += sizeof(T) * stride_a * batch;
-      size += sizeof(T) * stride_b * batch;
+      size += GetSizeA();
+      size += GetSizeB();
     }
     return size;
   }
@@ -167,13 +191,13 @@ struct GemmStridedBatchedParams : OpParams {
     *copy = *this;
     c10::DeviceIndex device = 0;
     AT_CUDA_CHECK(c10::cuda::GetDevice(&device));
-    size_t c_size = batch * stride_c * sizeof(T);
+    size_t c_size = GetSizeC();
     copy->c = static_cast<T*>(c10::cuda::CUDACachingAllocator::raw_alloc(c_size));
     AT_CUDA_CHECK(c10::cuda::CUDACachingAllocator::memcpyAsync(
         copy->c, device, c, device, c_size, getCurrentCUDAStream(device), true));
     if (duplicate_inputs) {
-      size_t a_size = sizeof(T) * stride_a * batch;
-      size_t b_size = sizeof(T) * stride_b * batch;
+      size_t a_size = GetSizeA();
+      size_t b_size = GetSizeB();
       copy->a = static_cast<const T*>(c10::cuda::CUDACachingAllocator::raw_alloc(a_size));
       copy->b = static_cast<const T*>(c10::cuda::CUDACachingAllocator::raw_alloc(b_size));
       copy->duplicate_inputs_ = true;
@@ -226,11 +250,23 @@ struct ScaledGemmParams : OpParams {
     return c10::str(transa, transb, "_", m, "_", n, "_", k);
   }
 
+  size_t GetSizeA() const {
+    return sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m);
+  }
+
+  size_t GetSizeB() const {
+    return sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k);
+  }
+
+  size_t GetSizeC() const {
+    return sizeof(T) * ldc * n;
+  }
+
   size_t GetSize(bool duplicate_inputs) const {
-    size_t size = sizeof(T) * ldc * n;
+    size_t size = GetSizeC();
     if (duplicate_inputs) {
-      size += sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m);
-      size += sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k);
+      size += GetSizeA();
+      size += GetSizeB();
     }
     return size;
   }
@@ -240,13 +276,13 @@ struct ScaledGemmParams : OpParams {
     *copy = *this;
     c10::DeviceIndex device = 0;
     AT_CUDA_CHECK(c10::cuda::GetDevice(&device));
-    size_t c_size = ldc * n * sizeof(T);
+    size_t c_size = GetSizeC();
     copy->c = c10::cuda::CUDACachingAllocator::raw_alloc(c_size);
     AT_CUDA_CHECK(c10::cuda::CUDACachingAllocator::memcpyAsync(
         copy->c, device, c, device, c_size, getCurrentCUDAStream(device), true));
     if (duplicate_inputs) {
-      size_t a_size = sizeof(T) * lda * ((transa == 'n' || transa == 'N') ? k : m);
-      size_t b_size = sizeof(T) * ldb * ((transb == 'n' || transb == 'N') ? n : k);
+      size_t a_size = GetSizeA();
+      size_t b_size = GetSizeB();
       copy->a = c10::cuda::CUDACachingAllocator::raw_alloc(a_size);
       copy->b = c10::cuda::CUDACachingAllocator::raw_alloc(b_size);
       copy->duplicate_inputs_ = true;

aten/src/ATen/cuda/tunable/Tunable.cpp

@@ -375,9 +375,9 @@ void TuningContext::EnableNumericsCheck(bool value) {
 }
 
 bool TuningContext::IsNumericsCheckEnabled() const {
-  static const char *env = getenv("PYTORCH_TUNABLEOP_NUMERICAL_CHECK");
-  if (env != nullptr && strcmp(env, "0") == 0) {
-    return false;
+  const char *env = getenv("PYTORCH_TUNABLEOP_NUMERICAL_CHECK");
+  if (env != nullptr && strcmp(env, "1") == 0) {
+    return true;
   }
   return numerics_check_enable_;
 }

aten/src/ATen/cuda/tunable/TunableOp.h

@@ -124,8 +124,11 @@ class TunableOp {
       std::string id_name = "Default";
       ParamsT* reference_params = nullptr;
 
+      // numeric check option is controlled by non-static env var, so check it once per tuned operator
+      bool do_numerics_check = ctx->IsNumericsCheckEnabled();
+
       // calcaulte a reference answer for numerical check
-      if (ctx->IsNumericsCheckEnabled()) {
+      if (do_numerics_check) {
         reference_params = params->DeepCopy(false);
         TORCH_CHECK(ops_[ResultEntry::Default()]->Call(reference_params) == OK);
       }
@@ -156,10 +159,11 @@ class TunableOp {
       for (size_t i = 0; i < op_names_.size(); i++) {
         auto* candidate = ops_[op_names_[i]].get(); // borrow pointer
 
-        if (ctx->IsNumericsCheckEnabled()) {
+        if (do_numerics_check) {
           ParamsT* numerical_params = params->DeepCopy(false);
           auto status = candidate->Call(numerical_params);
           if (status != OK) {
+            numerical_params->Delete();
             TUNABLE_LOG3("├──unsupported id=", i, ", ", op_sig, '(', params_sig, ") ", op_names_[i]);
             continue;
           }

aten/src/ATen/functorch/BatchRulesUnaryOps.cpp

@@ -59,13 +59,6 @@ view_as_complex_batch_rule(const Tensor& self, optional<int64_t> self_bdim) {
   return std::make_tuple(result, 0);
 }
 
-std::tuple<Tensor,optional<int64_t>>
-to_other_batch_rule(const Tensor& self, optional<int64_t> self_bdim,
-                    const Tensor& other, optional<int64_t> other_bdim,
-                    bool non_blocking,
-                    bool copy, std::optional<at::MemoryFormat> memory_format) {
-  return std::make_tuple(self.to(other, non_blocking, copy, memory_format), self_bdim);
-}
 }
 
 TORCH_LIBRARY_IMPL(aten, FuncTorchBatched, m) {

aten/src/ATen/native/Activation.h

@@ -23,7 +23,7 @@ enum class GeluType {
   END
 };
 
-static GeluType get_gelutype_enum(const c10::string_view approximate) {
+inline GeluType get_gelutype_enum(const c10::string_view approximate) {
   if (approximate == "none") {
     return GeluType::None;
   } else if (approximate == "tanh") {
@@ -33,7 +33,7 @@ static GeluType get_gelutype_enum(const c10::string_view approximate) {
   }
 }
 
-static std::string gelutype_to_string(const GeluType type) {
+inline std::string gelutype_to_string(const GeluType type) {
   switch(type) {
     case GeluType::None: return "none";
     case GeluType::Tanh: return "tanh";

aten/src/ATen/native/AdaptivePooling.h

@@ -28,15 +28,15 @@ using adaptive_max_pooling3d_backward_fn = void(*)(const Tensor& grad_input, con
 DECLARE_DISPATCH(adaptive_max_pooling3d_fn, adaptive_max_pool3d_kernel);
 DECLARE_DISPATCH(adaptive_max_pooling3d_backward_fn, adaptive_max_pool3d_backward_kernel);
 
-static inline int64_t start_index(int64_t a, int64_t b, int64_t c) {
+inline int64_t start_index(int64_t a, int64_t b, int64_t c) {
   return (a / b) * c + ((a % b) * c) / b;
 }
 
-static inline int64_t end_index(int64_t a, int64_t b, int64_t c) {
+inline int64_t end_index(int64_t a, int64_t b, int64_t c) {
   return 1 + ((a + 1) * c - 1) / b;
 }
 
-static inline void adaptive_pool_empty_output_check(const Tensor& gradOutput_, const char* arg_name) {
+inline void adaptive_pool_empty_output_check(const Tensor& gradOutput_, const char* arg_name) {
   int64_t ndim = gradOutput_.ndimension();
   for (const auto i : c10::irange(1, ndim)) {
     TORCH_CHECK(gradOutput_.size(i) > 0,

aten/src/ATen/native/ConvUtils.h

@@ -75,7 +75,7 @@ namespace {
   }
 }
 
-static inline bool cudnnv8_enabled_check_debug() {
+inline bool cudnnv8_enabled_check_debug() {
   static bool cudnnv8_flag = c10::utils::check_env("TORCH_CUDNN_V8_API_DISABLED") != true;
   static bool cudnnv8_debug = c10::utils::check_env("TORCH_CUDNN_V8_API_DEBUG") == true;
   static uint8_t cudnnv8_debugcount = 0;
@@ -86,7 +86,7 @@ static inline bool cudnnv8_enabled_check_debug() {
   return cudnnv8_flag == 1;
 }
 
-static inline bool cudnnv8_use_heur_mode_b() {
+inline bool cudnnv8_use_heur_mode_b() {
   return is_cudnnv8_heuristic_mode_b();
 }
 
@@ -186,7 +186,7 @@ static void check_args(CheckedFrom c, IntArrayRef args, size_t expected_size, co
 // (which the user can change) and computed inputs (which the user can
 // only indirectly affect).  It would be an interesting exercise to
 // come up with a general framework to handle such situations.)
-static void convolution_shape_check(
+inline void convolution_shape_check(
     CheckedFrom c,
     const TensorGeometryArg& input, const TensorGeometryArg& weight, const TensorGeometryArg& output,
     IntArrayRef padding, IntArrayRef stride, IntArrayRef dilation, int64_t groups)
@@ -212,7 +212,7 @@ static void convolution_shape_check(
 // takes an extra output_padding argument to resolve the ambiguity.
 
 template <typename T>
-static inline std::vector<T> _conv_output_size(
+inline std::vector<T> _conv_output_size(
     ArrayRef<T> input_size, ArrayRef<T> weight_size,
     ArrayRef<T> padding, ArrayRef<T> stride, ArrayRef<T> dilation = ArrayRef<T>()
 ) {
@@ -231,14 +231,14 @@ static inline std::vector<T> _conv_output_size(
   return output_size;
 }
 
-static inline std::vector<int64_t> conv_output_size(
+inline std::vector<int64_t> conv_output_size(
     IntArrayRef input_size, IntArrayRef weight_size,
     IntArrayRef padding, IntArrayRef stride, IntArrayRef dilation = IntArrayRef()
 ) {
   return _conv_output_size(input_size, weight_size, padding, stride, dilation);
 }
 
-static inline std::vector<c10::SymInt> conv_output_size(
+inline std::vector<c10::SymInt> conv_output_size(
     SymIntArrayRef input_size, SymIntArrayRef weight_size,
     SymIntArrayRef padding, SymIntArrayRef stride, SymIntArrayRef dilation = SymIntArrayRef()
 ) {
@@ -264,14 +264,14 @@ std::vector<T> _conv_input_size(
   return input_size;
 }
 
-static inline std::vector<c10::SymInt> conv_input_size(
+inline std::vector<c10::SymInt> conv_input_size(
     SymIntArrayRef output_size, SymIntArrayRef weight_size,
     SymIntArrayRef padding, SymIntArrayRef output_padding, SymIntArrayRef stride, SymIntArrayRef dilation, c10::SymInt groups
 ) {
   return _conv_input_size(output_size, weight_size, padding, output_padding, stride, dilation, groups);
 }
 
-static inline std::vector<int64_t> conv_input_size(
+inline std::vector<int64_t> conv_input_size(
     IntArrayRef output_size, IntArrayRef weight_size,
     IntArrayRef padding, IntArrayRef output_padding, IntArrayRef stride, IntArrayRef dilation, int64_t groups
 ) {
@@ -295,27 +295,27 @@ std::vector<T> _conv_weight_size(
   return weight_size;
 }
 
-static inline std::vector<c10::SymInt> conv_weight_size(
+inline std::vector<c10::SymInt> conv_weight_size(
     SymIntArrayRef input_size, SymIntArrayRef output_size,
     SymIntArrayRef padding, SymIntArrayRef output_padding, IntArrayRef stride, IntArrayRef dilation, int64_t groups
 ) {
   return _conv_weight_size(input_size, output_size, padding, output_padding, stride, dilation, groups);
 }
 
-static inline std::vector<int64_t> conv_weight_size(
+inline std::vector<int64_t> conv_weight_size(
     IntArrayRef input_size, IntArrayRef output_size,
     IntArrayRef padding, IntArrayRef output_padding, IntArrayRef stride, IntArrayRef dilation, int64_t groups
 ) {
   return _conv_weight_size(input_size, output_size, padding, output_padding, stride, dilation, groups);
 }
 
-static inline Tensor reshape_bias(int64_t dim, const Tensor& bias) {
+inline Tensor reshape_bias(int64_t dim, const Tensor& bias) {
   std::vector<int64_t> shape(dim, 1);
   shape[1] = -1;
   return bias.reshape(shape);
 }
 
-static inline at::MemoryFormat cudnn_conv_suggest_memory_format(const at::Tensor& input, const at::Tensor& weight) {
+inline at::MemoryFormat cudnn_conv_suggest_memory_format(const at::Tensor& input, const at::Tensor& weight) {
   // disable NHWC for float64 input.
   if (!at::detail::getCUDAHooks().compiledWithCuDNN() ||
       input.scalar_type() == at::kDouble ||
@@ -351,7 +351,7 @@ TORCH_API void _cudnn_set_conv_benchmark_empty_cache(bool enable);
 TORCH_API bool _cudnn_get_conv_benchmark_empty_cache();
 
 
-static inline bool miopen_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
+inline bool miopen_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
 
   // disable NHWC for float64 input.
   if (!at::detail::getCUDAHooks().compiledWithMIOpen() ||
@@ -378,7 +378,7 @@ static inline bool miopen_conv_use_channels_last(const at::Tensor& input, const
   return can_use_miopen_channels_last_2d || can_use_miopen_channels_last_3d;
 }
 
-static inline bool mkldnn_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
+inline bool mkldnn_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
 
   // disable NHWC for float64 input.
   if (input.scalar_type() == at::kDouble ||
@@ -405,7 +405,7 @@ static inline bool mkldnn_conv_use_channels_last(const at::Tensor& input, const
   return can_use_mkldnn_channels_last_2d || can_use_mkldnn_channels_last_3d;
 }
 
-static inline bool thnn_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
+inline bool thnn_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
 
   auto input_memory_format = input.suggest_memory_format();
   auto weight_memory_format = weight.suggest_memory_format();
@@ -417,7 +417,7 @@ static inline bool thnn_conv_use_channels_last(const at::Tensor& input, const at
   return can_use_thnn_channels_last_2d;
 }
 
-static inline bool xpu_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
+inline bool xpu_conv_use_channels_last(const at::Tensor& input, const at::Tensor& weight) {
 
   // check layout only for xpu tensor.
   if (!input.is_xpu() || !weight.is_xpu()) {

aten/src/ATen/native/Distributions.h

@@ -254,7 +254,7 @@ C10_DEVICE scalar_t sample_binomial(scalar_t count, scalar_t prob, BaseSampler<a
  * See note [3-Clause BSD License for the Cephes Math Library] in ATen/native/Math.h.
  */
 template<typename scalar_t, typename accscalar_t>
-C10_DEVICE static inline scalar_t digamma_one(scalar_t x) {
+C10_DEVICE inline scalar_t digamma_one(scalar_t x) {
   constexpr accscalar_t PSI_10 = 2.25175258906672110764;
   if (x == 0) {
     return INFINITY;
@@ -376,7 +376,7 @@ C10_HOST_DEVICE scalar_t standard_gamma_grad_one(scalar_t alpha_, scalar_t x_) {
 // Approximate reparameterized gradient of Beta(x,alpha,beta) wrt alpha.
 // Assumes x is close to zero and uses a Taylor expansion.
 template <typename scalar_t, typename accscalar_t>
-C10_DEVICE static inline scalar_t _beta_grad_alpha_small(scalar_t x, scalar_t alpha, scalar_t beta) {
+C10_DEVICE inline scalar_t _beta_grad_alpha_small(scalar_t x, scalar_t alpha, scalar_t beta) {
   const scalar_t factor = digamma_one<scalar_t, accscalar_t>(alpha)
                         - digamma_one<scalar_t, accscalar_t>(alpha + beta) - compat_log(x);
   scalar_t numer = 1;
@@ -394,7 +394,7 @@ C10_DEVICE static inline scalar_t _beta_grad_alpha_small(scalar_t x, scalar_t al
 // Approximate reparameterized gradient of Beta(x,alpha,beta) wrt beta.
 // Assumes x is close to zero and uses a Taylor expansion.
 template <typename scalar_t, typename accscalar_t>
-C10_DEVICE static inline scalar_t _beta_grad_beta_small(scalar_t x, scalar_t alpha, scalar_t beta) {
+C10_DEVICE inline scalar_t _beta_grad_beta_small(scalar_t x, scalar_t alpha, scalar_t beta) {
   const scalar_t factor = digamma_one<scalar_t, accscalar_t>(alpha + beta) - digamma_one<scalar_t, accscalar_t>(beta);
   scalar_t numer = 1, betas = 1, dbetas = 0, series = factor / alpha;
   for (int i = 1; i <= 8; ++i) {
@@ -412,7 +412,7 @@ C10_DEVICE static inline scalar_t _beta_grad_beta_small(scalar_t x, scalar_t alp
 // Assumes alpha and beta are both large and uses a Rice saddle point expansion.
 // To ensure numerical stability, this computation is performed at higher precision.
 template<typename scalar_t, typename accscalar_t>
-C10_DEVICE static inline scalar_t _beta_grad_alpha_mid(accscalar_t x, accscalar_t alpha, accscalar_t beta) {
+C10_DEVICE inline scalar_t _beta_grad_alpha_mid(accscalar_t x, accscalar_t alpha, accscalar_t beta) {
   const accscalar_t total = alpha + beta;
   const accscalar_t mean = alpha / total;
   const accscalar_t std = compat_sqrt(alpha * beta / (total + 1)) / total;
@@ -452,7 +452,7 @@ C10_DEVICE static inline scalar_t _beta_grad_alpha_mid(accscalar_t x, accscalar_
 // This function inputs total=alpha+beta to make it easy to implement
 // Dirichlet reparameterized gradients in terms of Betas.
 template<typename scalar_t, typename accscalar_t>
-C10_HOST_DEVICE static inline scalar_t dirichlet_grad_one(scalar_t x, scalar_t alpha, scalar_t total) {
+C10_HOST_DEVICE inline scalar_t dirichlet_grad_one(scalar_t x, scalar_t alpha, scalar_t total) {
   accscalar_t x_ = static_cast<accscalar_t>(x);
   accscalar_t alpha_ = static_cast<accscalar_t>(alpha);
   accscalar_t total_ = static_cast<accscalar_t>(total);

aten/src/ATen/native/FractionalMaxPooling.h

@@ -6,7 +6,7 @@
 namespace at::native {
 
 template<typename scalar_t>
-static inline std::vector<int> generate_intervals(
+inline std::vector<int> generate_intervals(
     scalar_t sample,
     int64_t inputSize,
     int64_t outputSize,
@@ -28,7 +28,7 @@ static inline std::vector<int> generate_intervals(
 }
 
 template <int64_t ndim>
-static inline void fractional_max_pool_check_shape(
+inline void fractional_max_pool_check_shape(
     const Tensor& input,
     const Tensor& randomSamples) {
 

aten/src/ATen/native/LinearAlgebraUtils.h

@@ -27,7 +27,7 @@
 
 namespace at::native {
 
-static inline c10::MaybeOwned<Tensor> expect_resolved_conj(const Tensor& tensor) {
+inline c10::MaybeOwned<Tensor> expect_resolved_conj(const Tensor& tensor) {
   if (tensor.is_conj()) {
     return c10::MaybeOwned<Tensor>::owned(tensor.resolve_conj());
   } else {
@@ -35,7 +35,7 @@ static inline c10::MaybeOwned<Tensor> expect_resolved_conj(const Tensor& tensor)
   }
 }
 
-static inline DimVector batched_matrix_contiguous_strides(
+inline DimVector batched_matrix_contiguous_strides(
     const IntArrayRef sizes,
     const bool f_contig = false) {
   // f_contig chooses between the strides of a batch of Fortran (F-contiguous)
@@ -62,7 +62,7 @@ static inline DimVector batched_matrix_contiguous_strides(
  *   P.data_ptr()[B * M * N] is of the same corresponding batch as the M' by N'
  *   matrix starting at Q.data_ptr()[B * M' * N'].
  */
-static inline Tensor cloneBatchedColumnMajor(const Tensor& src) {
+inline Tensor cloneBatchedColumnMajor(const Tensor& src) {
   // If src is already in batched column major format, then
   // this will be efficient (no reordering of the data will occur)
   // because the first transpose will make the tensor contiguous,
@@ -75,7 +75,7 @@ static inline Tensor cloneBatchedColumnMajor(const Tensor& src) {
 /*
  * contig chooses between C-contig (true) and F-contig (false)
  */
-static inline c10::MaybeOwned<Tensor> borrow_else_clone(const bool cond, const Tensor& borrow, const Tensor& clone, const bool contig) {
+inline c10::MaybeOwned<Tensor> borrow_else_clone(const bool cond, const Tensor& borrow, const Tensor& clone, const bool contig) {
   return cond ? c10::MaybeOwned<Tensor>::borrowed(borrow)
               : c10::MaybeOwned<Tensor>::owned(contig ? clone.clone(MemoryFormat::Contiguous)
                                                       : cloneBatchedColumnMajor(clone));
@@ -92,7 +92,7 @@ static inline c10::MaybeOwned<Tensor> borrow_else_clone(const bool cond, const T
  *  which is either the original batch size of the input, or its larger
  *  broadcasted shape.
  */
-static inline Tensor copyBatchedColumnMajor(const Tensor& src, int64_t nrows = -1,
+inline Tensor copyBatchedColumnMajor(const Tensor& src, int64_t nrows = -1,
     at::OptionalIntArrayRef desired_batch_sizes = c10::nullopt) {
   nrows = (nrows == -1) ? src.size(-2) : nrows;
   auto copy_sizes = desired_batch_sizes.has_value()
@@ -109,7 +109,7 @@ static inline Tensor copyBatchedColumnMajor(const Tensor& src, int64_t nrows = -
  * Given batches of matrices with arbitrary batch dim,
  * computes the number of batches.
  */
-static inline int64_t batchCount(const Tensor& batched_matrices) {
+inline int64_t batchCount(const Tensor& batched_matrices) {
   int64_t result = 1;
   for (int64_t i = 0; i < batched_matrices.ndimension() - 2; i++) {
     result *= batched_matrices.size(i);
@@ -118,15 +118,15 @@ static inline int64_t batchCount(const Tensor& batched_matrices) {
 }
 
 // Computes the number of elements of a matrix in a batched matrix tensor
-static inline int64_t matrixStride(const Tensor& batched_matrices) {
+inline int64_t matrixStride(const Tensor& batched_matrices) {
   return batched_matrices.size(-1) * batched_matrices.size(-2);
 }
 
 // Validates input shapes for operations on batches of square matrices (inverse, cholesky, symeig, eig)
-static inline void checkIsMatrix(const Tensor& A, const char* const f_name, const char* const arg_name = "A") {
+inline void checkIsMatrix(const Tensor& A, const char* const f_name, const char* const arg_name = "A") {
   TORCH_CHECK(A.dim() >= 2, f_name, ": The input tensor ", arg_name, " must have at least 2 dimensions.");
 }
-static inline void squareCheckInputs(const Tensor& self, const char* const f_name, const char* const arg_name = "A") {
+inline void squareCheckInputs(const Tensor& self, const char* const f_name, const char* const arg_name = "A") {
   checkIsMatrix(self, f_name, arg_name);
   TORCH_CHECK(self.sym_size(-1) == self.sym_size(-2),
               f_name,
@@ -134,7 +134,7 @@ static inline void squareCheckInputs(const Tensor& self, const char* const f_nam
               "but they are ", self.sym_size(-2), " by ", self.sym_size(-1), " matrices");
 }
 
-static inline void checkInputsSolver(const Tensor& A,
+inline void checkInputsSolver(const Tensor& A,
                                      const Tensor& B,
                                      const bool left,
                                      const char* const f_name) {
@@ -146,14 +146,14 @@ static inline void checkInputsSolver(const Tensor& A,
               " (", A.size(-2), "x", A.size(-1), " and ", B.size(-2), "x", B.size(-1), ")");
 }
 
-static inline bool is_row_or_column_contiguous(const Tensor& t) {
+inline bool is_row_or_column_contiguous(const Tensor& t) {
   // This could be made more general, similar to how it's checked in matmul, which would allow to
   // ellide the copy with strides such as (6, 12, 1, 3) or (3, 1, 9), but this is quite tricky.
   // We choose to be conservative for simplicity
   return t.is_contiguous() || t.transpose(-2, -1).is_contiguous();
 }
 
-static inline TransposeType to_transpose_type(const bool contig, const bool conj) {
+inline TransposeType to_transpose_type(const bool contig, const bool conj) {
   if (conj) {
     if (contig) { TORCH_INTERNAL_ASSERT(false, "Invalid transpose type"); }
     else {        return TransposeType::ConjTranspose; }
@@ -261,7 +261,7 @@ void batch_iterator_with_broadcasting(const Tensor& a, const Tensor& b, const fu
 }
 
 // Returns the epsilon value for floating types except half
-static inline double _get_epsilon(const ScalarType& sc_type) {
+inline double _get_epsilon(const ScalarType& sc_type) {
   switch (sc_type) {
     case at::ScalarType::Float:
       return static_cast<double>(std::numeric_limits<float>::epsilon());
@@ -274,7 +274,7 @@ static inline double _get_epsilon(const ScalarType& sc_type) {
 
 // Validates input shapes and devices
 // for linear solve methods (solve, cholesky_solve, lu_solve, triangular_solve)
-static inline void linearSolveCheckInputs(const Tensor& self, const Tensor& A, const char* name) {
+inline void linearSolveCheckInputs(const Tensor& self, const Tensor& A, const char* name) {
   TORCH_CHECK(self.device() == A.device(),
               "Expected b and A to be on the same device, but found b on ",
               self.device(), " and A on ", A.device(), " instead.");
@@ -293,7 +293,7 @@ static inline void linearSolveCheckInputs(const Tensor& self, const Tensor& A, c
               " but each b matrix is ", self.size(-2), " by ", self.size(-1));
 }
 
-static inline void checkFloatingOrComplex(const Tensor& t, const char* const f_name, const bool allow_low_precision_dtypes=true) {
+inline void checkFloatingOrComplex(const Tensor& t, const char* const f_name, const bool allow_low_precision_dtypes=true) {
   auto dtype = t.scalar_type();
   TORCH_CHECK((at::isFloatingType(dtype) || at::isComplexType(dtype)),
               f_name, ": Expected a floating point or complex tensor as input. Got ", dtype);
@@ -305,13 +305,13 @@ static inline void checkFloatingOrComplex(const Tensor& t, const char* const f_n
 
 
 // Checks if all the Tensors in a TensorList are of the same dimensions
-static inline void checkAllSameDim(TensorList tensors, int64_t dim) {
+inline void checkAllSameDim(TensorList tensors, int64_t dim) {
   for (auto &t : tensors) {
     TORCH_CHECK(t.dim() == dim, "Tensor dimension is ", t.dim(), ", expected ", dim, " instead.");
   }
 }
 
-static inline std::tuple<std::vector<int64_t>, std::vector<int64_t>> _linalg_broadcast_batch_dims(const Tensor& arg1, const Tensor& arg2) {
+inline std::tuple<std::vector<int64_t>, std::vector<int64_t>> _linalg_broadcast_batch_dims(const Tensor& arg1, const Tensor& arg2) {
   // broadcast the batch dimensions of arg1 and arg2.
   IntArrayRef arg1_batch_sizes(arg1.sizes().data(), arg1.ndimension() - 2);
   IntArrayRef arg2_batch_sizes(arg2.sizes().data(), arg2.ndimension() - 2);
@@ -325,7 +325,7 @@ static inline std::tuple<std::vector<int64_t>, std::vector<int64_t>> _linalg_bro
   return std::make_tuple(std::move(arg1_expand_size), std::move(arg2_expand_size));
 }
 
-static inline std::tuple<Tensor,Tensor> _linalg_broadcast_batch_dims(const Tensor& arg1, const Tensor& arg2, const char* name) {
+inline std::tuple<Tensor,Tensor> _linalg_broadcast_batch_dims(const Tensor& arg1, const Tensor& arg2, const char* name) {
   // If there's no name we assume we don't want to check the errors
   if (name != nullptr) {
     linearSolveCheckInputs(arg1, arg2, name);
@@ -338,7 +338,7 @@ static inline std::tuple<Tensor,Tensor> _linalg_broadcast_batch_dims(const Tenso
   return std::make_tuple(arg1_broadcasted, arg2_broadcasted);
 }
 
-static inline std::vector<int64_t> broadcast_batch_size(const Tensor& t1, const Tensor& t2, int64_t n_batch_dims) {
+inline std::vector<int64_t> broadcast_batch_size(const Tensor& t1, const Tensor& t2, int64_t n_batch_dims) {
   IntArrayRef t1_batch_sizes(t1.sizes().data(), n_batch_dims);
   IntArrayRef t2_batch_sizes(t2.sizes().data(), n_batch_dims);
   auto broadcasted_batch_sizes = infer_size(t1_batch_sizes, t2_batch_sizes);
@@ -346,7 +346,7 @@ static inline std::vector<int64_t> broadcast_batch_size(const Tensor& t1, const
 }
 
 // Return a permutation with the given axes moved to the end.
-static inline Tensor _move_to_end(const Tensor& self, IntArrayRef axes) {
+inline Tensor _move_to_end(const Tensor& self, IntArrayRef axes) {
   const std::vector<int64_t> a = axes.vec();
   const int64_t ndim = self.ndimension();
   std::vector<int64_t> perm;
@@ -368,7 +368,7 @@ static inline Tensor _move_to_end(const Tensor& self, IntArrayRef axes) {
 }
 
 // parse the "mode" param in linalg_qr: return a tuple of bools (compute_q, reduced)
-static inline std::tuple<bool, bool> _parse_qr_mode(c10::string_view mode) {
+inline std::tuple<bool, bool> _parse_qr_mode(c10::string_view mode) {
   bool compute_q;
   bool reduced;
   if (mode == "reduced") {
@@ -388,7 +388,7 @@ static inline std::tuple<bool, bool> _parse_qr_mode(c10::string_view mode) {
 }
 
 // Function to compute sizes, strides and the extra columns for the Q matrix in the QR Decomposition
-static inline std::tuple<DimVector, DimVector, int64_t> _compute_geometry_for_Q(
+inline std::tuple<DimVector, DimVector, int64_t> _compute_geometry_for_Q(
     const Tensor& input,
     bool reduced) {
   int64_t m = input.size(-2), n = input.size(-1);
@@ -407,7 +407,7 @@ static inline std::tuple<DimVector, DimVector, int64_t> _compute_geometry_for_Q(
   return std::make_tuple(q_sizes, q_strides, n_columns_q);
 }
 
-static inline bool svd_uses_cusolver(const Tensor& A) {
+inline bool svd_uses_cusolver(const Tensor& A) {
   // if cusolver is available, it is used unconditionally
   return A.is_cuda()
          && at::globalContext().hasCuSOLVER()
@@ -417,7 +417,7 @@ static inline bool svd_uses_cusolver(const Tensor& A) {
 
 // Function used instead of .to so that the original strides are retained
 // .to doesn't retain strides and make the output tensor contiguous
-static inline Tensor same_stride_to(const Tensor& original_tensor, const at::TensorOptions& options) {
+inline Tensor same_stride_to(const Tensor& original_tensor, const at::TensorOptions& options) {
   auto strided_to = at::empty_strided(original_tensor.sizes(),
                                       original_tensor.strides(),
                                       options);
@@ -433,7 +433,7 @@ static inline Tensor same_stride_to(const Tensor& original_tensor, const at::Ten
 // For instance, given a 4-D tensor, dimensions 1 and 3 can be shifted to the end by
 // calling `create_dim_backshift_permutation(1, 3, 4)`. The resulting vector will
 // be `vec(0, 2, 1, 3)`.
-static inline std::vector<int64_t> create_dim_backshift_permutation(int64_t dim0, int64_t dim1, int64_t ndim) {
+inline std::vector<int64_t> create_dim_backshift_permutation(int64_t dim0, int64_t dim1, int64_t ndim) {
   TORCH_CHECK(
     (dim0 != dim1) && (dim0 < ndim) && (dim0 >= 0) && (dim1 < ndim) && (dim1 >= 0),
     "duplicate or invalid dimensions");
@@ -453,7 +453,7 @@ static inline std::vector<int64_t> create_dim_backshift_permutation(int64_t dim0
 // will reverse a given permutation.
 // The reverse permutation array is created by swapping the indices and their
 // associated values from the given permutation array.
-static inline std::vector<int64_t> create_reverse_permutation(std::vector<int64_t> permutation) {
+inline std::vector<int64_t> create_reverse_permutation(std::vector<int64_t> permutation) {
   int64_t ndim = permutation.size();
   std::vector<int64_t> reverse_permutation(ndim);
   for (const auto dim_ind : c10::irange(ndim)) {
@@ -464,7 +464,7 @@ static inline std::vector<int64_t> create_reverse_permutation(std::vector<int64_
 
 // Compute R-work array size for MAGMA/LAPACK cgesdd/zgesdd
 // See https://github.com/Reference-LAPACK/lapack/blob/122506cd8b6ce050a200920c3d4c0b153b150fd8/SRC/cgesdd.f#L186
-static inline int64_t computeLRWorkDim(const char jobz, int64_t m, int64_t n) {
+inline int64_t computeLRWorkDim(const char jobz, int64_t m, int64_t n) {
   auto mn = std::min(m, n);
   auto mx = std::max(m, n);
   if (jobz == 'N') {
@@ -484,14 +484,14 @@ static inline int64_t computeLRWorkDim(const char jobz, int64_t m, int64_t n) {
 
 // This function checks whether the uplo argument input is valid
 // Allowed strings are "u", "U", "l", "L"
-static inline void checkUplo(const c10::string_view uplo) {
+inline void checkUplo(const c10::string_view uplo) {
   // To use std::toupper safely with plain chars (or signed chars), the argument should first be converted to unsigned char
   char uplo_uppercase = static_cast<char>(std::toupper(static_cast<unsigned char>(uplo[0])));
   TORCH_CHECK(uplo.size() == 1 && (uplo_uppercase == 'U' || uplo_uppercase == 'L'),
     "Expected UPLO argument to be 'L' or 'U', but got ", uplo);
 }
 
-static inline void checkSameDevice(const std::string& fn_name, Tensor result, Tensor input, const std::string& result_name = "result") {
+inline void checkSameDevice(const std::string& fn_name, Tensor result, Tensor input, const std::string& result_name = "result") {
   TORCH_CHECK(
       result.device() == input.device(),
       fn_name,
@@ -504,7 +504,7 @@ static inline void checkSameDevice(const std::string& fn_name, Tensor result, Te
 // (either floating or complex type input), so we can check whether input's dtype can be casted to result's dtype.
 // According to https://github.com/pytorch/pytorch/wiki/Developer-FAQ#how-does-out-work-in-pytorch
 // c10::canCast is used for checking the "safe copy" dtype requirements.
-static inline void checkLinalgCompatibleDtype(const std::string& fn_name, Tensor result, Tensor input, const std::string& result_name = "result") {
+inline void checkLinalgCompatibleDtype(const std::string& fn_name, Tensor result, Tensor input, const std::string& result_name = "result") {
   bool can_cast = c10::canCast(input.scalar_type(), result.scalar_type());
   TORCH_CHECK(
       can_cast,
@@ -514,7 +514,7 @@ static inline void checkLinalgCompatibleDtype(const std::string& fn_name, Tensor
 }
 
 // Alternatively, we can check whether the specific expected output type (result_type) can be safely casted to out tensor dtype (out_type)
-static inline void checkLinalgCompatibleDtype(const std::string& fn_name, ScalarType out_type, ScalarType result_type, const std::string& out_name = "result") {
+inline void checkLinalgCompatibleDtype(const std::string& fn_name, ScalarType out_type, ScalarType result_type, const std::string& out_name = "result") {
   bool can_cast = c10::canCast(result_type, out_type);
   TORCH_CHECK(
       can_cast,
@@ -523,7 +523,7 @@ static inline void checkLinalgCompatibleDtype(const std::string& fn_name, Scalar
       out_name, " with dtype ", out_type);
 }
 
-static inline void checkNotComplexTolerance(const Tensor& tol, const c10::string_view f_name, const c10::string_view tol_name) {
+inline void checkNotComplexTolerance(const Tensor& tol, const c10::string_view f_name, const c10::string_view tol_name) {
   TORCH_CHECK(!at::isComplexType(tol.scalar_type()),
               f_name, ": ", tol_name, " tensor of complex type is not supported. Got ", tol.scalar_type());
 }
@@ -538,7 +538,7 @@ static inline void checkNotComplexTolerance(const Tensor& tol, const c10::string
   Let input.shape = (batch_dimensions, m, n), then 'other' is of vector type if other.shape == (batch_dimensions, m).
   This rule is compatible with NumPy, see https://github.com/numpy/numpy/blob/v1.20.0/numpy/linalg/linalg.py#L384-L389
 */
-static inline bool linalg_solve_is_vector_rhs(const Tensor& input, const Tensor& other) {
+inline bool linalg_solve_is_vector_rhs(const Tensor& input, const Tensor& other) {
   auto expected_batched_rhs_shape = SymIntArrayRef(input.sym_sizes().data(), input.dim() - 1); // input.shape[:-1]
   bool vector_case = other.dim() == 1 || (input.dim() - 1 == other.dim() && other.sym_sizes().equals(expected_batched_rhs_shape));
   return vector_case;
@@ -547,7 +547,7 @@ static inline bool linalg_solve_is_vector_rhs(const Tensor& input, const Tensor&
 /*
   Computes linear indices for a tensor with original_shape to access its elements like it was a materialized broadcast tensor.
 */
-static inline Tensor get_linear_indices(int64_t numel, IntArrayRef original_shape, IntArrayRef broadcast_shape) {
+inline Tensor get_linear_indices(int64_t numel, IntArrayRef original_shape, IntArrayRef broadcast_shape) {
   TensorOptions options = at::TensorOptions().dtype(at::kLong).device(at::kCPU);
   return at::arange(numel, options).view(original_shape).broadcast_to(broadcast_shape).contiguous();
 }
@@ -578,7 +578,7 @@ class BroadcastLinearIndices {
   }
 };
 
-static inline bool is_blas_compatible_column_major_order(const Tensor& input) {
+inline bool is_blas_compatible_column_major_order(const Tensor& input) {
   IntArrayRef input_strides = input.strides();
   IntArrayRef input_sizes = input.sizes();
   auto ndim = input.dim();
@@ -599,7 +599,7 @@ static inline bool is_blas_compatible_column_major_order(const Tensor& input) {
       batch_stride_compatible;
 }
 
-static inline bool is_blas_compatible_row_major_order(const Tensor& input) {
+inline bool is_blas_compatible_row_major_order(const Tensor& input) {
   IntArrayRef input_strides = input.strides();
   IntArrayRef input_sizes = input.sizes();
   auto ndim = input.dim();