Revert "[Release 2.4] Release only changes - use pinned triton." (#129139)

Revert "[Release 2.4] Release only changes - use pinned triton. (#128388)"

This reverts commit 1cd41997e99ae1722be3fe88e1867af5f6779433.

.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-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)"

.github/ci_commit_pins/audio.txt

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

.github/ci_commit_pins/xla.txt

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

.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/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

@@ -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
 
@@ -60,13 +60,11 @@ 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 }}
 
       - name: Build Triton wheel
-        env:
-          IS_RELEASE_TAG: ${{ startsWith(github.event.ref, 'refs/tags/v') }}
         run: |
           set -x
           mkdir -p "${RUNNER_TEMP}/artifacts/"
@@ -107,14 +105,9 @@ jobs:
             BUILD_ROCM="--build-rocm"
           fi
 
-          RELEASE=""
-          if [[ "${IS_RELEASE_TAG}" == true ]]; then
-            RELEASE="--release"
-          fi
-
           docker exec -t "${container_name}" yum install -y zlib-devel zip
           docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}"  -m pip install -U setuptools==67.4.0
-          docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}" /pytorch/.github/scripts/build_triton_wheel.py $BUILD_ROCM $RELEASE
+          docker exec -t "${container_name}" "${PYTHON_EXECUTABLE}" /pytorch/.github/scripts/build_triton_wheel.py $BUILD_ROCM
           docker exec -t "${container_name}" chown -R 1000.1000 /artifacts
 
       - uses: actions/upload-artifact@v3
@@ -124,7 +117,7 @@ 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 +202,12 @@ 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
 
@@ -222,13 +215,11 @@ 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 }}
 
       - name: Build Triton conda package
-        env:
-          IS_RELEASE_TAG: ${{ startsWith(github.event.ref, 'refs/tags/v') }}
         run: |
           set -x
           mkdir -p "${RUNNER_TEMP}/artifacts/"
@@ -241,13 +232,8 @@ jobs:
             "${DOCKER_IMAGE}" \
           )
 
-          RELEASE=""
-          if [[ "${IS_RELEASE_TAG}" == true ]]; then
-            RELEASE="--release"
-          fi
-
           docker exec -t "${container_name}" yum install -y llvm11 llvm11-devel llvm11-static llvm11-libs zlib-devel
-          docker exec -t "${container_name}" python /pytorch/.github/scripts/build_triton_wheel.py --build-conda --py-version="${PY_VERS}" $RELEASE
+          docker exec -t "${container_name}" python /pytorch/.github/scripts/build_triton_wheel.py --build-conda --py-version="${PY_VERS}"
           docker exec -t "${container_name}" chown -R 1000.1000 /artifacts
 
       - uses: actions/upload-artifact@v3
@@ -257,7 +243,7 @@ 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/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-binary-manywheel-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
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu
       build_environment: linux-binary-manywheel
@@ -66,7 +66,7 @@ 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: manywheel-py3_8-cpu
       build_environment: linux-binary-manywheel
@@ -87,7 +87,7 @@ 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: manywheel-py3_8-cpu
     secrets:
@@ -107,7 +107,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-cxx11-abi
@@ -126,7 +126,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-cxx11-abi
@@ -148,7 +148,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-cpu-cxx11-abi
@@ -170,7 +170,7 @@ 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
@@ -190,7 +190,7 @@ 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
@@ -212,7 +212,7 @@ 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
     secrets:
@@ -233,7 +233,7 @@ 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
@@ -253,7 +253,7 @@ 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
@@ -275,7 +275,7 @@ 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
     secrets:
@@ -296,7 +296,7 @@ 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
@@ -316,7 +316,7 @@ 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
@@ -338,7 +338,7 @@ 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
     secrets:
@@ -359,7 +359,7 @@ 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
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-rocm6_0
       build_environment: linux-binary-manywheel
@@ -380,7 +380,7 @@ 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
       DESIRED_PYTHON: "3.8"
     steps:
       - name: Setup ROCm
@@ -393,7 +393,6 @@ 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
@@ -405,7 +404,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -421,7 +420,7 @@ 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
@@ -441,7 +440,7 @@ 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
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-rocm6_0
     secrets:
@@ -462,7 +461,7 @@ 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
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-rocm6_1
       build_environment: linux-binary-manywheel
@@ -483,7 +482,7 @@ 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
       DESIRED_PYTHON: "3.8"
     steps:
       - name: Setup ROCm
@@ -496,7 +495,6 @@ 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
@@ -508,7 +506,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -524,7 +522,7 @@ 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
@@ -544,7 +542,7 @@ 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
       DESIRED_PYTHON: "3.8"
       build_name: manywheel-py3_8-rocm6_1
     secrets:
@@ -564,7 +562,7 @@ 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: manywheel-py3_9-cpu
       build_environment: linux-binary-manywheel
@@ -582,7 +580,7 @@ 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: manywheel-py3_9-cpu
       build_environment: linux-binary-manywheel
@@ -603,7 +601,7 @@ 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: manywheel-py3_9-cpu
     secrets:
@@ -623,7 +621,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-cxx11-abi
@@ -642,7 +640,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-cxx11-abi
@@ -664,7 +662,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-cpu-cxx11-abi
@@ -686,7 +684,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda11_8
       build_environment: linux-binary-manywheel
@@ -706,7 +704,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda11_8
       build_environment: linux-binary-manywheel
@@ -728,7 +726,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda11_8
     secrets:
@@ -749,7 +747,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda12_1
       build_environment: linux-binary-manywheel
@@ -769,7 +767,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda12_1
       build_environment: linux-binary-manywheel
@@ -791,7 +789,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda12_1
     secrets:
@@ -812,7 +810,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda12_4
       build_environment: linux-binary-manywheel
@@ -832,7 +830,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda12_4
       build_environment: linux-binary-manywheel
@@ -854,7 +852,7 @@ 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: "3.9"
       build_name: manywheel-py3_9-cuda12_4
     secrets:
@@ -875,7 +873,7 @@ 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
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-rocm6_0
       build_environment: linux-binary-manywheel
@@ -896,7 +894,7 @@ 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
       DESIRED_PYTHON: "3.9"
     steps:
       - name: Setup ROCm
@@ -909,7 +907,6 @@ 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
@@ -921,7 +918,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -937,7 +934,7 @@ 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
@@ -957,7 +954,7 @@ 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
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-rocm6_0
     secrets:
@@ -978,7 +975,7 @@ 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
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-rocm6_1
       build_environment: linux-binary-manywheel
@@ -999,7 +996,7 @@ 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
       DESIRED_PYTHON: "3.9"
     steps:
       - name: Setup ROCm
@@ -1012,7 +1009,6 @@ 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
@@ -1024,7 +1020,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1040,7 +1036,7 @@ 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
@@ -1060,7 +1056,7 @@ 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
       DESIRED_PYTHON: "3.9"
       build_name: manywheel-py3_9-rocm6_1
     secrets:
@@ -1080,7 +1076,7 @@ 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: manywheel-py3_10-cpu
       build_environment: linux-binary-manywheel
@@ -1098,7 +1094,7 @@ 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: manywheel-py3_10-cpu
       build_environment: linux-binary-manywheel
@@ -1119,7 +1115,7 @@ 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: manywheel-py3_10-cpu
     secrets:
@@ -1139,7 +1135,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-cxx11-abi
@@ -1158,7 +1154,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-cxx11-abi
@@ -1180,7 +1176,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-cpu-cxx11-abi
@@ -1202,7 +1198,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda11_8
       build_environment: linux-binary-manywheel
@@ -1222,7 +1218,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda11_8
       build_environment: linux-binary-manywheel
@@ -1244,7 +1240,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda11_8
     secrets:
@@ -1265,7 +1261,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda12_1
       build_environment: linux-binary-manywheel
@@ -1285,7 +1281,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda12_1
       build_environment: linux-binary-manywheel
@@ -1307,7 +1303,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda12_1
     secrets:
@@ -1328,7 +1324,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda12_4
       build_environment: linux-binary-manywheel
@@ -1348,7 +1344,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda12_4
       build_environment: linux-binary-manywheel
@@ -1370,7 +1366,7 @@ 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: "3.10"
       build_name: manywheel-py3_10-cuda12_4
     secrets:
@@ -1391,7 +1387,7 @@ 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
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-rocm6_0
       build_environment: linux-binary-manywheel
@@ -1412,7 +1408,7 @@ 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
       DESIRED_PYTHON: "3.10"
     steps:
       - name: Setup ROCm
@@ -1425,7 +1421,6 @@ 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
@@ -1437,7 +1432,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1453,7 +1448,7 @@ 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
@@ -1473,7 +1468,7 @@ 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
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-rocm6_0
     secrets:
@@ -1494,7 +1489,7 @@ 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
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-rocm6_1
       build_environment: linux-binary-manywheel
@@ -1515,7 +1510,7 @@ 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
       DESIRED_PYTHON: "3.10"
     steps:
       - name: Setup ROCm
@@ -1528,7 +1523,6 @@ 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
@@ -1540,7 +1534,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1556,7 +1550,7 @@ 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
@@ -1576,7 +1570,7 @@ 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
       DESIRED_PYTHON: "3.10"
       build_name: manywheel-py3_10-rocm6_1
     secrets:
@@ -1596,7 +1590,7 @@ 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: manywheel-py3_11-cpu
       build_environment: linux-binary-manywheel
@@ -1614,7 +1608,7 @@ 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: manywheel-py3_11-cpu
       build_environment: linux-binary-manywheel
@@ -1635,7 +1629,7 @@ 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: manywheel-py3_11-cpu
     secrets:
@@ -1655,7 +1649,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-cxx11-abi
@@ -1674,7 +1668,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-cxx11-abi
@@ -1696,7 +1690,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-cpu-cxx11-abi
@@ -1718,7 +1712,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda11_8
       build_environment: linux-binary-manywheel
@@ -1738,7 +1732,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda11_8
       build_environment: linux-binary-manywheel
@@ -1760,7 +1754,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda11_8
     secrets:
@@ -1781,7 +1775,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda12_1
       build_environment: linux-binary-manywheel
@@ -1801,7 +1795,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda12_1
       build_environment: linux-binary-manywheel
@@ -1823,7 +1817,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda12_1
     secrets:
@@ -1844,7 +1838,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda12_4
       build_environment: linux-binary-manywheel
@@ -1864,7 +1858,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda12_4
       build_environment: linux-binary-manywheel
@@ -1886,7 +1880,7 @@ 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: "3.11"
       build_name: manywheel-py3_11-cuda12_4
     secrets:
@@ -1907,7 +1901,7 @@ 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
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-rocm6_0
       build_environment: linux-binary-manywheel
@@ -1928,7 +1922,7 @@ 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
       DESIRED_PYTHON: "3.11"
     steps:
       - name: Setup ROCm
@@ -1941,7 +1935,6 @@ 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
@@ -1953,7 +1946,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -1969,7 +1962,7 @@ 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
@@ -1989,7 +1982,7 @@ 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
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-rocm6_0
     secrets:
@@ -2010,7 +2003,7 @@ 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
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-rocm6_1
       build_environment: linux-binary-manywheel
@@ -2031,7 +2024,7 @@ 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
       DESIRED_PYTHON: "3.11"
     steps:
       - name: Setup ROCm
@@ -2044,7 +2037,6 @@ 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
@@ -2056,7 +2048,7 @@ jobs:
       - name: Checkout pytorch/builder
         uses: malfet/checkout@silent-checkout
         with:
-          ref: main
+          ref: release/2.4
           submodules: recursive
           repository: pytorch/builder
           path: builder
@@ -2072,7 +2064,7 @@ 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
@@ -2092,7 +2084,7 @@ 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
       DESIRED_PYTHON: "3.11"
       build_name: manywheel-py3_11-rocm6_1
     secrets:
@@ -2112,7 +2104,7 @@ 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: manywheel-py3_12-cpu
       build_environment: linux-binary-manywheel
@@ -2130,7 +2122,7 @@ 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: manywheel-py3_12-cpu
       build_environment: linux-binary-manywheel
@@ -2151,7 +2143,7 @@ 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: manywheel-py3_12-cpu
     secrets:
@@ -2171,7 +2163,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-cxx11-abi
@@ -2190,7 +2182,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-cxx11-abi
@@ -2212,7 +2204,7 @@ jobs:
       #       favor of GPU_ARCH_VERSION
       DESIRED_CUDA: cpu-cxx11-abi
       GPU_ARCH_TYPE: cpu-cxx11-abi
-      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-main
+      DOCKER_IMAGE: pytorch/manylinuxcxx11-abi-builder:cpu-cxx11-abi-2.4
       DESIRED_DEVTOOLSET: cxx11-abi
       DESIRED_PYTHON: "3.12"
       build_name: manywheel-py3_12-cpu-cxx11-abi
@@ -2234,7 +2226,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda11_8
       build_environment: linux-binary-manywheel
@@ -2254,7 +2246,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda11_8
       build_environment: linux-binary-manywheel
@@ -2276,7 +2268,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda11_8
     secrets:
@@ -2297,7 +2289,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda12_1
       build_environment: linux-binary-manywheel
@@ -2317,7 +2309,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda12_1
       build_environment: linux-binary-manywheel
@@ -2339,7 +2331,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda12_1
     secrets:
@@ -2360,7 +2352,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda12_4
       build_environment: linux-binary-manywheel
@@ -2380,7 +2372,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda12_4
       build_environment: linux-binary-manywheel
@@ -2402,7 +2394,7 @@ 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: "3.12"
       build_name: manywheel-py3_12-cuda12_4
     secrets:

.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/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/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();

aten/src/ATen/native/LossNLL.cpp

@@ -675,15 +675,6 @@ Tensor nll_loss_symint(const Tensor & self, const Tensor & target, const std::op
   return std::get<0>(at::nll_loss_forward_symint(self, target, weight, reduction, std::move(ignore_index)));
 }
 
-// Duplicate of above code for non-symbolic ints. Kept for BC purposes and to minimize breakages.
-static Tensor nll_loss(const Tensor & self, const Tensor & target, const std::optional<Tensor>& weight_opt, int64_t reduction, int64_t ignore_index) {
-  // See [Note: hacky wrapper removal for optional tensor]
-  c10::MaybeOwned<Tensor> weight_maybe_owned = at::borrow_from_optional_tensor(weight_opt);
-  const Tensor& weight = *weight_maybe_owned;
-
-  return std::get<0>(at::nll_loss_forward_symint(self, target, weight, reduction, ignore_index));
-}
-
 Tensor nll_loss_nd_symint(
     const Tensor& self,
     const Tensor& target,

aten/src/ATen/native/Math.h

@@ -147,7 +147,7 @@ jiterator_also_stringify_as(jiterator_code(
 #define CENTRAL_RANGE 0.7
 
 template <typename T>
-static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 calc_erfinv(T y) {
 /* Function to calculate inverse error function.  Rational approximation
 is used to generate an initial approximation, which is then improved to
@@ -232,7 +232,7 @@ Date:  February 1996
  * See note [3-Clause BSD License for the Cephes Math Library].
  */
 template <typename scalar_t, bool is_cuda=false>
-C10_HOST_DEVICE static inline scalar_t zeta(scalar_t x, scalar_t q) __ubsan_ignore_float_divide_by_zero__ {
+C10_HOST_DEVICE inline scalar_t zeta(scalar_t x, scalar_t q) __ubsan_ignore_float_divide_by_zero__ {
   using acc_t = at::acc_type<scalar_t, is_cuda>;
   const acc_t MACHEP = acc_t{1.11022302462515654042E-16};
   constexpr acc_t zero = acc_t{0.0};
@@ -324,7 +324,7 @@ C10_HOST_DEVICE static inline scalar_t zeta(scalar_t x, scalar_t q) __ubsan_igno
  *            N                   0
  */
 template <typename T>
-C10_HOST_DEVICE static inline T polevl(const T x, const T A[], size_t len) {
+C10_HOST_DEVICE inline T polevl(const T x, const T A[], size_t len) {
   T result = 0;
   for (size_t i = 0; i <= len; i++) {
     result = result * x + A[i];
@@ -332,7 +332,7 @@ C10_HOST_DEVICE static inline T polevl(const T x, const T A[], size_t len) {
   return result;
 }
 
-static inline double trigamma(double x) __ubsan_ignore_float_divide_by_zero__ {
+inline double trigamma(double x) __ubsan_ignore_float_divide_by_zero__ {
   double sign = +1;
   double result = 0;
   if (x < 0.5) {
@@ -350,7 +350,7 @@ static inline double trigamma(double x) __ubsan_ignore_float_divide_by_zero__ {
   return sign * result;
 }
 
-static inline float trigamma(float x) __ubsan_ignore_float_divide_by_zero__ {
+inline float trigamma(float x) __ubsan_ignore_float_divide_by_zero__ {
   float sign = +1;
   float result = 0;
   if (x < 0.5f) {
@@ -372,7 +372,7 @@ static inline float trigamma(float x) __ubsan_ignore_float_divide_by_zero__ {
  * This function is derived from the implementation of the digamma function in the Cephes Math Library.
  * See note [3-Clause BSD License for the Cephes Math Library].
  */
-static inline double calc_digamma(double x) {
+inline double calc_digamma(double x) {
   // [C++ Standard Reference: Gamma Function] https://en.cppreference.com/w/cpp/numeric/math/tgamma
   static double PSI_10 = 2.25175258906672110764;
   if (x == 0) {
@@ -430,7 +430,7 @@ static inline double calc_digamma(double x) {
  * This function is derived from the implementation of the digamma function in the Cephes Math Library.
  * See note [3-Clause BSD License for the Cephes Math Library].
  */
-static inline float calc_digamma(float x) {
+inline float calc_digamma(float x) {
   // See [C++ Standard Reference: Gamma Function]
   static float PSI_10 = 2.25175258906672110764f;
   if (x == 0) {
@@ -485,16 +485,16 @@ static inline float calc_digamma(float x) {
   return result + logf(x) - (0.5f / x) - y;
 }
 
-static inline c10::BFloat16 calc_digamma(c10::BFloat16 a) {
+inline c10::BFloat16 calc_digamma(c10::BFloat16 a) {
   return calc_digamma(static_cast<float>(a));
 }
 
-static inline c10::Half calc_digamma(c10::Half a) {
+inline c10::Half calc_digamma(c10::Half a) {
   return calc_digamma(static_cast<float>(a));
 }
 
 template <typename scalar_t, bool is_cuda=false>
-static inline C10_HOST_DEVICE scalar_t calc_polygamma(scalar_t x, int n) {
+inline C10_HOST_DEVICE scalar_t calc_polygamma(scalar_t x, int n) {
   // already blocked if n <= 1
   const auto one = scalar_t{1};
   return ((n % 2) ? one : -one) *
@@ -519,7 +519,7 @@ static inline C10_HOST_DEVICE scalar_t calc_polygamma(scalar_t x, int n) {
  * See NOTICE for the licenses.
  */
 template <typename scalar_t>
-static scalar_t ratevl(scalar_t x, const scalar_t num[], int64_t M,
+scalar_t ratevl(scalar_t x, const scalar_t num[], int64_t M,
     const scalar_t denom[], int64_t N) {
   // evaluating rational function, i.e., the ratio of two polynomials
   // the coefficients for numerator are given by `num` while coeffs for
@@ -1061,7 +1061,7 @@ static scalar_t _igamc_helper_continued_fraction(scalar_t a, scalar_t x) {
 }
 
 template <typename scalar_t>
-static inline scalar_t calc_igammac(scalar_t a, scalar_t x) {
+inline scalar_t calc_igammac(scalar_t a, scalar_t x) {
   /* the calculation of the regularized upper incomplete gamma function
    * is done differently based on the values of a and x:
    * - if x and/or a is at the boundary of defined region, then assign the
@@ -1141,7 +1141,7 @@ static inline scalar_t calc_igammac(scalar_t a, scalar_t x) {
 }
 
 template <typename scalar_t>
-static inline scalar_t calc_igamma(scalar_t a, scalar_t x) {
+scalar_t calc_igamma(scalar_t a, scalar_t x) {
   /* the calculation of the regularized lower incomplete gamma function
    * is done differently based on the values of a and x:
    * - if x and/or a is at the boundary of defined region, then assign the
@@ -1203,39 +1203,39 @@ static inline scalar_t calc_igamma(scalar_t a, scalar_t x) {
 }
 
 template <>
-C10_UNUSED c10::BFloat16 calc_igamma<c10::BFloat16>(c10::BFloat16 a, c10::BFloat16 x) {
+C10_UNUSED inline c10::BFloat16 calc_igamma<c10::BFloat16>(c10::BFloat16 a, c10::BFloat16 x) {
   return calc_igamma<float>(float(a), float(x));
 }
 
 template <>
-C10_UNUSED c10::Half calc_igamma<c10::Half>(c10::Half a, c10::Half x) {
+C10_UNUSED inline c10::Half calc_igamma<c10::Half>(c10::Half a, c10::Half x) {
   return calc_igamma<float>(float(a), float(x));
 }
 
 template <>
-C10_UNUSED c10::BFloat16 calc_igammac<c10::BFloat16>(c10::BFloat16 a, c10::BFloat16 x) {
+C10_UNUSED inline c10::BFloat16 calc_igammac<c10::BFloat16>(c10::BFloat16 a, c10::BFloat16 x) {
   return calc_igammac<float>(float(a), float(x));
 }
 
 template <>
-C10_UNUSED c10::Half calc_igammac<c10::Half>(c10::Half a, c10::Half x) {
+C10_UNUSED inline c10::Half calc_igammac<c10::Half>(c10::Half a, c10::Half x) {
   return calc_igammac<float>(float(a), float(x));
 }
 
 inline c10::BFloat16 calc_erfinv(c10::BFloat16 a) { return calc_erfinv(float(a)); }
 
 template <typename T>
-static T abs_impl(T v) {
+inline T abs_impl(T v) {
   return std::abs(v);
 }
 
 template <>
-C10_UNUSED uint8_t abs_impl(uint8_t v) {
+C10_UNUSED inline uint8_t abs_impl(uint8_t v) {
   return v;
 }
 
 template <typename T>
-static inline typename std::enable_if<std::is_integral<T>::value, T>::type
+inline typename std::enable_if<std::is_integral<T>::value, T>::type
 calc_gcd(T a, T b) {
   a = abs_impl(a);
   b = abs_impl(b);
@@ -1284,7 +1284,7 @@ C10_HOST_DEVICE c10::complex<T> exp2_impl(c10::complex<T> x) {
  * required is x -> 2(2ab/x - b - a)/(b-a).  If b is infinity, this becomes x -> 4a/x - 1.
  */
 template <typename T>
-static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 chbevl(const T x, const T array[], size_t len) {
   T b0, b1, b2;
 
@@ -1310,7 +1310,7 @@ chbevl(const T x, const T array[], size_t len) {
  * of all inputs to convert them into the domain of the approximation.
  */
 template <typename T>
-static inline std::tuple<const T*, size_t> chebyshev_coefficients_i0e_A() {
+inline std::tuple<const T*, size_t> chebyshev_coefficients_i0e_A() {
   /* Chebyshev coefficients for exp(-x) I0(x)
    * in the interval [0,8].
    *
@@ -1336,7 +1336,7 @@ static inline std::tuple<const T*, size_t> chebyshev_coefficients_i0e_A() {
 };
 
 template <typename T>
-static inline std::tuple<const T*, size_t> chebyshev_coefficients_i0e_B() {
+inline std::tuple<const T*, size_t> chebyshev_coefficients_i0e_B() {
   /* Chebyshev coefficients for exp(-x) sqrt(x) I0(x)
    * in the inverted interval [8,infinity].
    *
@@ -1361,7 +1361,7 @@ static inline std::tuple<const T*, size_t> chebyshev_coefficients_i0e_B() {
 };
 
 template <typename T>
-static inline typename std::enable_if<std::is_same<double, T>::value, std::tuple<const T*, size_t>>::type
+inline typename std::enable_if<std::is_same<double, T>::value, std::tuple<const T*, size_t>>::type
 chebyshev_coefficients_i1e_A() {
   /* Chebyshev coefficients for exp(-x) I1(x)
    * in the interval [0,8].
@@ -1388,7 +1388,7 @@ chebyshev_coefficients_i1e_A() {
 };
 
 template <typename T>
-static inline typename std::enable_if<std::is_same<float, T>::value, std::tuple<const T*, size_t>>::type
+inline typename std::enable_if<std::is_same<float, T>::value, std::tuple<const T*, size_t>>::type
 chebyshev_coefficients_i1e_A() {
   /* Chebyshev coefficients for exp(-x) I1(x)
    * in the interval [0,8].
@@ -1417,7 +1417,7 @@ chebyshev_coefficients_i1e_A() {
 };
 
 template <typename T>
-static inline typename std::enable_if<std::is_same<double, T>::value, std::tuple<const T*, size_t>>::type
+inline typename std::enable_if<std::is_same<double, T>::value, std::tuple<const T*, size_t>>::type
 chebyshev_coefficients_i1e_B() {
   /* Chebyshev coefficients for exp(-x) sqrt(x) I1(x)
    * in the inverted interval [8,infinity].
@@ -1443,7 +1443,7 @@ chebyshev_coefficients_i1e_B() {
 };
 
 template <typename T>
-static inline typename std::enable_if<std::is_same<float, T>::value, std::tuple<const T*, size_t>>::type
+inline typename std::enable_if<std::is_same<float, T>::value, std::tuple<const T*, size_t>>::type
 chebyshev_coefficients_i1e_B() {
   /* Chebyshev coefficients for exp(-x) sqrt(x) I1(x)
    * in the inverted interval [8,infinity].
@@ -1463,7 +1463,7 @@ chebyshev_coefficients_i1e_B() {
 };
 
 template <typename T>
-static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 calc_i0(T _x) {
   T x = std::abs(_x);
 
@@ -1481,7 +1481,7 @@ calc_i0(T _x) {
 }
 
 // Upcast bfloat16 input to float for numerical accuracy purposes
-static inline c10::BFloat16 calc_i0(c10::BFloat16 a) { return calc_i0(static_cast<float>(a)); }
+inline c10::BFloat16 calc_i0(c10::BFloat16 a) { return calc_i0(static_cast<float>(a)); }
 
 /*
  * This function is derived from the implementation of the i1 function in the Cephes Math Library.
@@ -1493,7 +1493,7 @@ static inline c10::BFloat16 calc_i0(c10::BFloat16 a) { return calc_i0(static_cas
  * of all inputs to convert them into the domain of the approximation.
  */
 template <typename T>
-static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 calc_i1(T _x) {
   T x = std::abs(_x);
 
@@ -1522,7 +1522,7 @@ calc_i1(T _x) {
  * of all inputs to convert them into the domain of the approximation.
  */
 template <typename T>
-static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 calc_i1e(T _x) {
   T x = std::abs(_x);
 
@@ -1549,7 +1549,7 @@ calc_i1e(T _x) {
  * (integrated from minus infinity to x) is equal to y.
  */
 template <typename T>
-static inline C10_HOST_DEVICE T calc_ndtri(T y0) {
+inline C10_HOST_DEVICE T calc_ndtri(T y0) {
 
   /* sqrt(2pi) */
   constexpr T s2pi = 2.50662827463100050242E0;
@@ -1737,7 +1737,7 @@ static inline C10_HOST_DEVICE T calc_ndtri(T y0) {
 
 
 template <typename T>
-C10_HOST_DEVICE  static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+C10_HOST_DEVICE  inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 erfcx_y100(T y100)
 {
   switch (static_cast<int>(y100)) {
@@ -2148,7 +2148,7 @@ return 0.97771701335885035464e0 + (0.22000938572830479551e-1 + (0.27951610702682
 }
 
 template <typename T>
-C10_HOST_DEVICE static inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
+C10_HOST_DEVICE inline typename std::enable_if<std::is_floating_point<T>::value, T>::type
 calc_erfcx(T x)
 {
   if (at::_isnan(x)) {
@@ -2188,7 +2188,7 @@ calc_erfcx(T x)
  * See NOTICE for the licenses.
  */
 template <typename T>
-static inline C10_HOST_DEVICE T calc_log_ndtr(T x) {
+inline C10_HOST_DEVICE T calc_log_ndtr(T x) {
   T t = x * c10::frac_sqrt_2<T>;
   if (x < T{-1.0}) {
     return std::log(calc_erfcx(-t) / 2) - t * t;
@@ -2198,7 +2198,7 @@ static inline C10_HOST_DEVICE T calc_log_ndtr(T x) {
 }
 
 template<typename T>
-static inline C10_HOST_DEVICE T airy_ai_forward(T x) {
+inline C10_HOST_DEVICE T airy_ai_forward(T x) {
     static const T AN[] = {
             +3.46538101525629032477e-01,
             +1.20075952739645805542e+01,
@@ -2377,7 +2377,7 @@ static inline C10_HOST_DEVICE T airy_ai_forward(T x) {
 } // T airy_ai(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T bessel_j0_forward(T x) {
+inline C10_HOST_DEVICE T bessel_j0_forward(T x) {
     static const T PP[] = {
             +7.96936729297347051624e-04,
             +8.28352392107440799803e-02,
@@ -2489,7 +2489,7 @@ static inline C10_HOST_DEVICE T bessel_j0_forward(T x) {
 } // bessel_j0_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T bessel_j1_forward(T x) {
+inline C10_HOST_DEVICE T bessel_j1_forward(T x) {
     static const T PP[] = {
             +7.62125616208173112003e-04,
             +7.31397056940917570436e-02,
@@ -2597,7 +2597,7 @@ static inline C10_HOST_DEVICE T bessel_j1_forward(T x) {
 } // bessel_j1_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T bessel_y0_forward(T x) {
+inline C10_HOST_DEVICE T bessel_y0_forward(T x) {
     static const T PP[] = {
             +7.96936729297347051624e-04,
             +8.28352392107440799803e-02,
@@ -2712,7 +2712,7 @@ static inline C10_HOST_DEVICE T bessel_y0_forward(T x) {
 } // bessel_y0_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T bessel_y1_forward(T x) {
+inline C10_HOST_DEVICE T bessel_y1_forward(T x) {
     static const T PP[] = {
             +7.62125616208173112003e-04,
             +7.31397056940917570436e-02,
@@ -2826,7 +2826,7 @@ static inline C10_HOST_DEVICE T bessel_y1_forward(T x) {
 } // bessel_y1_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_t_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_t_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -2865,12 +2865,12 @@ static inline C10_HOST_DEVICE T chebyshev_polynomial_t_forward(T x, int64_t n) {
 } // chebyshev_polynomial_t_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_t_forward(T x, T n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_t_forward(T x, T n) {
     return chebyshev_polynomial_t_forward(x, static_cast<int64_t>(n));
 } // chebyshev_polynomial_t_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_u_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_u_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -2913,12 +2913,12 @@ static inline C10_HOST_DEVICE T chebyshev_polynomial_u_forward(T x, int64_t n) {
 } // chebyshev_polynomial_u_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_u_forward(T x, T n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_u_forward(T x, T n) {
     return chebyshev_polynomial_u_forward(x, static_cast<int64_t>(n));
 } // chebyshev_polynomial_u_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_v_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_v_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -2969,12 +2969,12 @@ static inline C10_HOST_DEVICE T chebyshev_polynomial_v_forward(T x, int64_t n) {
 } // chebyshev_polynomial_v_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_v_forward(T x, T n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_v_forward(T x, T n) {
     return chebyshev_polynomial_v_forward(x, static_cast<int64_t>(n));
 } // chebyshev_polynomial_v_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_w_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_w_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3029,12 +3029,12 @@ static inline C10_HOST_DEVICE T chebyshev_polynomial_w_forward(T x, int64_t n) {
 } // chebyshev_polynomial_w_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T chebyshev_polynomial_w_forward(T x, T n) {
+inline C10_HOST_DEVICE T chebyshev_polynomial_w_forward(T x, T n) {
     return chebyshev_polynomial_w_forward(x, static_cast<int64_t>(n));
 } // chebyshev_polynomial_w_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3061,17 +3061,17 @@ static inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, int64_t n) {
 } // hermite_polynomial_h_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false, std::enable_if_t<!std::is_floating_point<T>::value, int> = 0>
-static inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, T n) {
+inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, T n) {
     return hermite_polynomial_h_forward(x, static_cast<int64_t>(n));
 } // hermite_polynomial_h_forward(T x, T n)
 
 template<typename T, bool is_cuda=false, std::enable_if_t<std::is_floating_point<T>::value, int> = 0>
-static inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, T n) {
+inline C10_HOST_DEVICE T hermite_polynomial_h_forward(T x, T n) {
     return hermite_polynomial_h_forward(x, ((!std::isinf(n)) && (!std::isnan(n))) ? static_cast<int64_t>(n) : static_cast<int64_t>(-1));
 } // hermite_polynomial_h_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T hermite_polynomial_he_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T hermite_polynomial_he_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3098,12 +3098,12 @@ static inline C10_HOST_DEVICE T hermite_polynomial_he_forward(T x, int64_t n) {
 } // hermite_polynomial_he_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T hermite_polynomial_he_forward(T x, T n) {
+inline C10_HOST_DEVICE T hermite_polynomial_he_forward(T x, T n) {
     return hermite_polynomial_he_forward(x, static_cast<int64_t>(n));
 } // hermite_polynomial_he_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T laguerre_polynomial_l_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T laguerre_polynomial_l_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3134,12 +3134,12 @@ static inline C10_HOST_DEVICE T laguerre_polynomial_l_forward(T x, int64_t n) {
 } // laguerre_polynomial_l_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T laguerre_polynomial_l_forward(T x, T n) {
+inline C10_HOST_DEVICE T laguerre_polynomial_l_forward(T x, T n) {
     return laguerre_polynomial_l_forward(x, static_cast<int64_t>(n));
 } // laguerre_polynomial_l_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T legendre_polynomial_p_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T legendre_polynomial_p_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3174,12 +3174,12 @@ static inline C10_HOST_DEVICE T legendre_polynomial_p_forward(T x, int64_t n) {
 } // legendre_polynomial_p_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T legendre_polynomial_p_forward(T x, T n) {
+inline C10_HOST_DEVICE T legendre_polynomial_p_forward(T x, T n) {
     return legendre_polynomial_p_forward(x, static_cast<int64_t>(n));
 } // legendre_polynomial_p_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T modified_bessel_i0_forward(T x) {
+inline C10_HOST_DEVICE T modified_bessel_i0_forward(T x) {
     static const T A[] = {
             -4.41534164647933937950e-18,
             +3.33079451882223809783e-17,
@@ -3268,7 +3268,7 @@ static inline C10_HOST_DEVICE T modified_bessel_i0_forward(T x) {
 } // modified_bessel_i0_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T modified_bessel_i1_forward(T x) {
+inline C10_HOST_DEVICE T modified_bessel_i1_forward(T x) {
     static const T A[] = {
             +2.77791411276104639959e-18,
             -2.11142121435816608115e-17,
@@ -3364,7 +3364,7 @@ static inline C10_HOST_DEVICE T modified_bessel_i1_forward(T x) {
 } // modified_bessel_i1_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T modified_bessel_k0_forward(T x) {
+inline C10_HOST_DEVICE T modified_bessel_k0_forward(T x) {
     static const T A[] = {
             +1.37446543561352307156e-16,
             +4.25981614279661018399e-14,
@@ -3441,7 +3441,7 @@ static inline C10_HOST_DEVICE T modified_bessel_k0_forward(T x) {
 } // modified_bessel_k0_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T modified_bessel_k1_forward(T x) {
+inline C10_HOST_DEVICE T modified_bessel_k1_forward(T x) {
     static const T A[] = {
             -7.02386347938628759343e-18,
             -2.42744985051936593393e-15,
@@ -3519,7 +3519,7 @@ static inline C10_HOST_DEVICE T modified_bessel_k1_forward(T x) {
 } // modified_bessel_k1_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T scaled_modified_bessel_k0_forward(T x) {
+inline C10_HOST_DEVICE T scaled_modified_bessel_k0_forward(T x) {
     static const T A[] = {
             +1.37446543561352307156e-16,
             +4.25981614279661018399e-14,
@@ -3596,7 +3596,7 @@ static inline C10_HOST_DEVICE T scaled_modified_bessel_k0_forward(T x) {
 } // T scaled_modified_bessel_k0_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T scaled_modified_bessel_k1_forward(T x) {
+inline C10_HOST_DEVICE T scaled_modified_bessel_k1_forward(T x) {
     static const T A[] = {
             -7.02386347938628759343e-18,
             -2.42744985051936593393e-15,
@@ -3674,7 +3674,7 @@ static inline C10_HOST_DEVICE T scaled_modified_bessel_k1_forward(T x) {
 } // T scaled_modified_bessel_k1_forward(T x)
 
 template<typename T>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_t_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_t_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3717,12 +3717,12 @@ static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_t_forward(T x, int6
 } // shifted_chebyshev_polynomial_t_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_t_forward(T x, T n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_t_forward(T x, T n) {
     return shifted_chebyshev_polynomial_t_forward(x, static_cast<int64_t>(n));
 } // shifted_chebyshev_polynomial_t_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_u_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_u_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3769,12 +3769,12 @@ static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_u_forward(T x, int6
 } // shifted_chebyshev_polynomial_u_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_u_forward(T x, T n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_u_forward(T x, T n) {
     return shifted_chebyshev_polynomial_u_forward(x, static_cast<int64_t>(n));
 } // shifted_chebyshev_polynomial_u_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_v_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_v_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3825,12 +3825,12 @@ static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_v_forward(T x, int6
 } // shifted_chebyshev_polynomial_v_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_v_forward(T x, T n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_v_forward(T x, T n) {
     return shifted_chebyshev_polynomial_v_forward(x, static_cast<int64_t>(n));
 } // shifted_chebyshev_polynomial_v_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_w_forward(T x, int64_t n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_w_forward(T x, int64_t n) {
     if (n < 0) {
         return T(0.0);
     }
@@ -3881,12 +3881,12 @@ static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_w_forward(T x, int6
 } // shifted_chebyshev_polynomial_w_forward(T x, int64_t n)
 
 template<typename T, bool is_cuda=false>
-static inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_w_forward(T x, T n) {
+inline C10_HOST_DEVICE T shifted_chebyshev_polynomial_w_forward(T x, T n) {
     return shifted_chebyshev_polynomial_w_forward(x, static_cast<int64_t>(n));
 } // shifted_chebyshev_polynomial_w_forward(T x, T n)
 
 template<typename T>
-static inline C10_HOST_DEVICE T spherical_bessel_j0_forward(T x) {
+inline C10_HOST_DEVICE T spherical_bessel_j0_forward(T x) {
     if (std::isinf(x)) {
         return T(0.0);
     }

aten/src/ATen/native/Padding.h

@@ -26,7 +26,7 @@ DECLARE_DISPATCH(padding_fn, replication_pad3d_backward_kernel);
 namespace padding {
 
 template <int dim>
-static inline void check_valid_input(const Tensor& input, IntArrayRef padding) {
+inline void check_valid_input(const Tensor& input, IntArrayRef padding) {
 
   TORCH_CHECK(padding.size() == 2 * dim,
       "padding size is expected to be ", 2 * dim,

aten/src/ATen/native/Pool.h

@@ -48,7 +48,7 @@ DECLARE_DISPATCH(max_pool3d_backward_fn, max_pool3d_backward_kernel);
 namespace {
 
 template <typename dest_t, typename src_t>
-static inline dest_t
+inline dest_t
 safe_downcast(src_t v)
 {
   TORCH_CHECK(std::numeric_limits<dest_t>::min() <= v && v <= std::numeric_limits<dest_t>::max(),
@@ -58,7 +58,7 @@ safe_downcast(src_t v)
 }
 
 template<typename T>
-static inline T pooling_output_shape_pad_lr(
+inline T pooling_output_shape_pad_lr(
         T inputSize, T kernelSize, T pad_l, T pad_r, T stride, T dilation,
         bool ceil_mode) {
     T outputSize = div_rtn<T>(
@@ -75,7 +75,7 @@ static inline T pooling_output_shape_pad_lr(
 }
 
 template<typename T>
-static inline T pooling_output_shape(
+inline T pooling_output_shape(
       T inputSize, T kernelSize, T pad, T stride, T dilation, bool ceil_mode) {
     TORCH_CHECK(stride != 0, "stride should not be zero");
     TORCH_CHECK(pad >= 0,
@@ -117,7 +117,7 @@ inline std::pair<c10::SymInt, c10::SymInt> pooling_same_mode_padding_lr(
 }
 
 // AveragePool2d/DilatedMaxPool2d (forward)
-static inline void
+inline void
 pool2d_shape_check(
   const Tensor& input,
   int kH, int kW, int dH, int dW, int padH, int padW, int dilationH, int dilationW,
@@ -164,7 +164,7 @@ pool2d_shape_check(
 }
 
 // DilatedMaxPool2d (backward)
-static inline void
+inline void
 max_pool2d_backward_shape_check(
   const Tensor& input,
   const Tensor& gradOutput,
@@ -192,7 +192,7 @@ max_pool2d_backward_shape_check(
 }
 
 // AveragePool2d (backward)
-static inline void
+inline void
 avg_pool2d_backward_shape_check(
   const Tensor& input,
   const Tensor& gradOutput,
@@ -218,7 +218,7 @@ avg_pool2d_backward_shape_check(
 }
 
 // AveragePool3d/DilatedMaxPool3d (forward)
-static inline void
+inline void
 pool3d_shape_check(
   const Tensor& input,
   int64_t nslices,
@@ -280,7 +280,7 @@ pool3d_shape_check(
               "Output size is too small");
 }
 
-static inline void
+inline void
 max_pool3d_backward_shape_check(
   const Tensor& input,
   const Tensor& gradOutput,
@@ -317,7 +317,7 @@ max_pool3d_backward_shape_check(
   check_dim_size(indices, ndim, ndim-1, owidth);
 }
 
-static inline void
+inline void
 avg_pool3d_backward_shape_check(
   const Tensor& input,
   const Tensor& gradOutput,

aten/src/ATen/native/Pow.h

@@ -24,7 +24,7 @@ namespace native {
 // only non-zero result.
 template <class T,
   typename std::enable_if<std::is_integral<T>::value, T>::type* = nullptr>
-static inline HOST_DEVICE __ubsan_ignore_signed_int_overflow__ T powi_impl(T a, T b) {
+inline HOST_DEVICE __ubsan_ignore_signed_int_overflow__ T powi_impl(T a, T b) {
   T result = 1;
   while (b) {
     if (b & 1) {
@@ -38,13 +38,13 @@ static inline HOST_DEVICE __ubsan_ignore_signed_int_overflow__ T powi_impl(T a,
 
 template <class T,
   typename std::enable_if<std::is_integral<T>::value && !std::is_signed<T>::value, T>::type* = nullptr>
-static inline HOST_DEVICE T powi(T a, T b) {
+inline HOST_DEVICE T powi(T a, T b) {
   return powi_impl(a, b);
 }
 
 template <class T,
   typename std::enable_if<std::is_integral<T>::value && std::is_signed<T>::value, T>::type* = nullptr>
-static inline HOST_DEVICE T powi(T a, T b) {
+inline HOST_DEVICE T powi(T a, T b) {
   if ( b < 0 ) {
       if ( a == 1 ) {
           return 1;

aten/src/ATen/native/ReduceOpsUtils.h

@@ -31,7 +31,7 @@ constexpr scalar_t lower_bound() {
   return lim::has_infinity ? -lim::infinity() : lim::lowest();
 }
 
-static inline Tensor restride_dim(
+inline Tensor restride_dim(
   const Tensor& src, int64_t dim,
   IntArrayRef replacement_shape
 ) {
@@ -96,13 +96,13 @@ inline std::optional<Tensor> _allreduce_return_trivial(
     " but found ", out.option())\
 }
 
-static inline void check_scalar_type_device_layout_equal(const Tensor& out, const Tensor& self) {
+inline void check_scalar_type_device_layout_equal(const Tensor& out, const Tensor& self) {
   OPTION_TYPE_EQUALITY_CHECK(scalar_type, out, self);
   OPTION_TYPE_EQUALITY_CHECK(device, out.options(), self.options());
   OPTION_TYPE_EQUALITY_CHECK(layout, out.options(), self.options());
 }
 
-static inline Tensor integer_upcast(const Tensor& self, std::optional<ScalarType> dtype) {
+inline Tensor integer_upcast(const Tensor& self, std::optional<ScalarType> dtype) {
   ScalarType scalarType = self.scalar_type();
   TORCH_CHECK(!isBarebonesUnsignedType(scalarType), "integer upcasting for uint16, uint32 and uint64 is not currently implemented");
   ScalarType upcast_scalarType = dtype.value_or(at::isIntegralType(scalarType, /*includeBool=*/true) ? ScalarType::Long : scalarType);
@@ -111,7 +111,7 @@ static inline Tensor integer_upcast(const Tensor& self, std::optional<ScalarType
 
 using DimMask = TensorIterator::DimMask;
 
-static DimVector make_dim_vector(OptionalIntArrayRef opt_dims, int64_t ndim) {
+inline DimVector make_dim_vector(OptionalIntArrayRef opt_dims, int64_t ndim) {
   if (opt_dims.has_value()) {
     return DimVector(opt_dims.value());
   } else {
@@ -121,7 +121,7 @@ static DimVector make_dim_vector(OptionalIntArrayRef opt_dims, int64_t ndim) {
   }
 }
 
-static DimMask make_dim_mask(OptionalIntArrayRef opt_dims, int64_t ndim, bool allow_empty_dims=false) {
+inline DimMask make_dim_mask(OptionalIntArrayRef opt_dims, int64_t ndim, bool allow_empty_dims=false) {
   DimMask mask;
   if (opt_dims.has_value()) {
     auto dims = opt_dims.value();
@@ -150,7 +150,7 @@ inline DimVector shape_from_dim_mask(const Tensor& self, DimMask mask, bool keep
   return shape;
 }
 
-static void resize_reduction_result(
+inline void resize_reduction_result(
     Tensor& result, const Tensor& self, DimMask mask, bool keepdim,
     ScalarType /*dtype*/)
 {
@@ -167,7 +167,7 @@ inline Tensor create_reduction_result(
   return at::empty(shape, self.options().dtype(dtype));
 }
 
-static Tensor review_reduce_result(const Tensor& result, int ndim, DimMask mask, bool keepdim) {
+inline Tensor review_reduce_result(const Tensor& result, int ndim, DimMask mask, bool keepdim) {
   if (keepdim) {
     return result;
   }
@@ -182,7 +182,7 @@ static Tensor review_reduce_result(const Tensor& result, int ndim, DimMask mask,
   return result.as_strided(shape, stride);
 }
 
-static TensorIterator make_reduction(
+inline TensorIterator make_reduction(
     const char* name, Tensor& result, const Tensor& self,
     at::OptionalIntArrayRef dim_opt,
     bool keepdim, ScalarType in_dtype, ScalarType out_dtype) {
@@ -207,7 +207,7 @@ static TensorIterator make_reduction(
   return TensorIterator::reduce_op(viewed_result, self.to(in_dtype));
 }
 
-static C10_UNUSED TensorIterator make_reduction(
+inline C10_UNUSED TensorIterator make_reduction(
     const char* name, Tensor& result, const Tensor& self,
     at::OptionalIntArrayRef dim, bool keepdim, ScalarType out_dtype) {
   // special case for type promotion in mixed precision, improves computational
@@ -222,7 +222,7 @@ static C10_UNUSED TensorIterator make_reduction(
   return make_reduction(name, result, self, dim, keepdim, in_dtype, out_dtype);
 }
 
-static TensorIterator make_reduction(
+inline TensorIterator make_reduction(
     const char* name, Tensor& result1, Tensor& result2, const Tensor& self,
     at::OptionalIntArrayRef dim_opt, bool keepdim, ScalarType dtype1,
     ScalarType dtype2) {
@@ -259,13 +259,13 @@ static TensorIterator make_reduction(
   return TensorIterator::reduce_op(viewed_result1, viewed_result2, self.to(dtype1));
 }
 
-static C10_UNUSED TensorIterator make_reduction(
+inline C10_UNUSED TensorIterator make_reduction(
     const char* name, Tensor& result1, Tensor& result2, const Tensor& self,
     at::OptionalIntArrayRef dim, bool keepdim, ScalarType dtype) {
   return make_reduction(name, result1, result2, self, dim, keepdim, dtype, dtype);
 }
 
-static void zero_numel_check_dims(const Tensor& self, const int64_t dim, const char *fn_name) {
+inline void zero_numel_check_dims(const Tensor& self, const int64_t dim, const char *fn_name) {
   if (self.ndimension() == 0) {
     TORCH_CHECK_INDEX(dim == 0 || dim == -1, fn_name,
       ": Expected reduction dim -1 or 0 for scalar but got ", dim);
@@ -276,7 +276,7 @@ static void zero_numel_check_dims(const Tensor& self, const int64_t dim, const c
   }
 }
 
-static void zero_numel_check_dims(const Tensor& self, const IntArrayRef dim, const char *fn_name) {
+inline void zero_numel_check_dims(const Tensor& self, const IntArrayRef dim, const char *fn_name) {
   TORCH_CHECK(
     !dim.empty(),
       fn_name, ": Expected reduction dim to be specified for input.numel() == 0. ",
@@ -286,7 +286,7 @@ static void zero_numel_check_dims(const Tensor& self, const IntArrayRef dim, con
   }
 }
 
-static std::vector<int64_t> get_zero_numel_tensor_size(
+inline std::vector<int64_t> get_zero_numel_tensor_size(
     const Tensor& self,
     const int64_t dim,
     const bool keepdim,
@@ -313,7 +313,7 @@ static std::vector<int64_t> get_zero_numel_tensor_size(
 // This function should be called when you are reducing a zero-numel tensor and want to
 // resize the output and return it. This function exists for resizing zero-numel
 // tensors when the size of the reduction dimension is non-zero.
-static C10_UNUSED void zero_numel_tensor_resize(Tensor& result, Tensor& result_indices,
+inline C10_UNUSED void zero_numel_tensor_resize(Tensor& result, Tensor& result_indices,
                                      const Tensor& self, const int64_t dim,
                                      const bool keepdim, const char *fn_name) {
   auto sizes = get_zero_numel_tensor_size(self, dim, keepdim, fn_name);
@@ -349,7 +349,7 @@ inline ScalarType get_dtype_from_result(Tensor& result, std::optional<ScalarType
 
 namespace at::meta {
 
-static C10_UNUSED DimVector get_reduction_shape(
+inline C10_UNUSED DimVector get_reduction_shape(
     const Tensor& self,
     IntArrayRef dims,
     bool keepdim,
@@ -358,7 +358,7 @@ static C10_UNUSED DimVector get_reduction_shape(
   return native::shape_from_dim_mask(self, mask, keepdim);
 }
 
-static void resize_reduction(
+inline void resize_reduction(
     impl::MetaBase& meta,
     const Tensor& self,
     OptionalIntArrayRef opt_dims,
@@ -379,7 +379,7 @@ static void resize_reduction(
       meta.maybe_get_output(), self, dims_, keepdim);
 }
 
-static void resize_reduction_with_indices(
+inline void resize_reduction_with_indices(
     impl::MetaBase& meta,
     const Tensor& self,
     IntArrayRef dims,
@@ -396,7 +396,7 @@ static void resize_reduction_with_indices(
       meta.maybe_get_output(1), self, dims_, keepdim);
 }
 
-static TensorIterator make_reduction(
+inline TensorIterator make_reduction(
     const Tensor& self,
     const Tensor& result,
     OptionalIntArrayRef opt_dims,
@@ -412,7 +412,7 @@ static TensorIterator make_reduction(
   return TensorIterator::reduce_op(viewed_result, self.to(in_dtype));
 }
 
-static TensorIterator make_reduction(
+inline TensorIterator make_reduction(
     const Tensor& self,
     const Tensor& result1,
     const Tensor& result2,
@@ -434,7 +434,7 @@ static TensorIterator make_reduction(
   return TensorIterator::reduce_op(viewed_result1, viewed_result2, self.to(dtype1));
 }
 
-static C10_UNUSED TensorIterator make_reduction_from_out_ty(
+inline C10_UNUSED TensorIterator make_reduction_from_out_ty(
     const Tensor& self,
     const Tensor& result,
     OptionalIntArrayRef opt_dims,

aten/src/ATen/native/ReductionType.h

@@ -6,7 +6,7 @@ namespace at::native {
 
 enum class ReductionType {MAX, MEAN, MIN, SUM, PROD};
 
-static inline ReductionType get_reduction_enum(const c10::string_view& reduce) {
+inline ReductionType get_reduction_enum(const c10::string_view& reduce) {
   if (reduce == "max" || reduce == "amax") {
     return ReductionType::MAX;
   } else if (reduce == "mean") {
@@ -23,7 +23,7 @@ static inline ReductionType get_reduction_enum(const c10::string_view& reduce) {
 }
 
 // used for `scatter_reduce`, old options for BC.
-static inline ReductionType get_operator_enum(const c10::string_view reduce, bool use_new_options) {
+inline ReductionType get_operator_enum(const c10::string_view reduce, bool use_new_options) {
   if (use_new_options) {
     return get_reduction_enum(reduce);
   } else {

aten/src/ATen/native/Resize.h

@@ -40,7 +40,7 @@ TORCH_API void resize_bytes_cpu(StorageImpl* storage, size_t size_bytes);
 TORCH_API void resize_bytes_meta(StorageImpl* storage, c10::SymInt size_bytes);
 TORCH_API void resize_bytes_nocuda(const Storage& storage, c10::SymInt size_bytes);
 
-static inline void maybe_resize_storage_cpu(TensorImpl* self, size_t new_size_bytes) {
+inline void maybe_resize_storage_cpu(TensorImpl* self, size_t new_size_bytes) {
   // It does not make sense to try to resize a storage
   // to hold 0 elements, and this can break
   // if storage_offset is positive but
@@ -79,7 +79,7 @@ template <>
 inline int64_t maybe_convert_symint(c10::SymInt x) { return x.guard_int(__FILE__, __LINE__); }
 
 template <typename T>
-static inline void checkInBoundsForStorage(
+inline void checkInBoundsForStorage(
     ArrayRef<T> size,
     ArrayRef<T> stride,
     T storage_offset,
@@ -111,7 +111,7 @@ static inline void checkInBoundsForStorage(
 }
 
 template <typename T>
-static inline void checkSetStorage(Tensor& result, Storage storage, T storage_offset,
+inline void checkSetStorage(Tensor& result, Storage storage, T storage_offset,
                                    ArrayRef<T> size, ArrayRef<T> stride) {
   // FIXME: stride should be optional
   if (stride.data()) {

aten/src/ATen/native/TriangularOps.cpp

@@ -180,10 +180,6 @@ TORCH_IMPL_FUNC(triu_cpu)(const Tensor& self, int64_t k, const Tensor &result) {
   compute_triu_tril<UpperTriangle>(self, k, result);
 }
 
-static Tensor trace_backward(const Tensor& grad, at::IntArrayRef sizes) {
-    return at::native::trace_backward_symint(grad, c10::fromIntArrayRefSlow(sizes));
-}
-
 Tensor trace_backward_symint(const Tensor& grad, c10::SymIntArrayRef sizes) {
   if (sizes.size() != 2) {
     throw std::runtime_error("expected matrix input");

aten/src/ATen/native/UpSample.h

@@ -210,7 +210,7 @@ std::array<int64_t, 5> upsample_3d_common_check(IntArrayRef input_size, IntArray
   return {nbatch, channels, output_depth, output_height, output_width};
 }
 
-static inline void upsample_2d_shape_check(
+inline void upsample_2d_shape_check(
     const Tensor& input,
     const Tensor& grad_output,
     int64_t nbatch,
@@ -251,7 +251,7 @@ static inline void upsample_2d_shape_check(
 }
 
 template <typename scalar_t>
-static inline scalar_t compute_scales_value(
+inline scalar_t compute_scales_value(
     const std::optional<double> scale,
     int64_t input_size,
     int64_t output_size) {
@@ -263,7 +263,7 @@ static inline scalar_t compute_scales_value(
 }
 
 template <typename scalar_t>
-static inline scalar_t area_pixel_compute_scale(
+inline scalar_t area_pixel_compute_scale(
     int64_t input_size,
     int64_t output_size,
     bool align_corners,
@@ -281,7 +281,7 @@ static inline scalar_t area_pixel_compute_scale(
 }
 
 template <typename scalar_t>
-static inline scalar_t area_pixel_compute_source_index(
+inline scalar_t area_pixel_compute_source_index(
     scalar_t scale,
     int64_t dst_index,
     bool align_corners,
@@ -308,7 +308,7 @@ static inline scalar_t area_pixel_compute_source_index(
   }
 }
 
-static inline int64_t nearest_neighbor_compute_source_index(
+inline int64_t nearest_neighbor_compute_source_index(
     const float scale,
     int64_t dst_index,
     int64_t input_size) {
@@ -319,7 +319,7 @@ static inline int64_t nearest_neighbor_compute_source_index(
   return src_index;
 }
 
-static inline int64_t nearest_neighbor_exact_compute_source_index(
+inline int64_t nearest_neighbor_exact_compute_source_index(
     const float scale,
     int64_t dst_index,
     int64_t input_size) {
@@ -331,7 +331,7 @@ static inline int64_t nearest_neighbor_exact_compute_source_index(
   return src_index;
 }
 
-static inline int64_t nearest_idx(
+inline int64_t nearest_idx(
     int64_t output_index,
     int64_t input_size,
     int64_t output_size,
@@ -352,7 +352,7 @@ static inline int64_t nearest_idx(
   }
 }
 
-static inline int64_t nearest_exact_idx(
+inline int64_t nearest_exact_idx(
     int64_t output_index,
     int64_t input_size,
     int64_t output_size,
@@ -392,17 +392,17 @@ static void upsample_increment_value_bounded(
 // Based on
 // https://en.wikipedia.org/wiki/Bicubic_interpolation#Bicubic_convolution_algorithm
 template <typename scalar_t>
-static inline scalar_t cubic_convolution1(scalar_t x, scalar_t A) {
+inline scalar_t cubic_convolution1(scalar_t x, scalar_t A) {
   return ((A + 2) * x - (A + 3)) * x * x + 1;
 }
 
 template <typename scalar_t>
-static inline scalar_t cubic_convolution2(scalar_t x, scalar_t A) {
+inline scalar_t cubic_convolution2(scalar_t x, scalar_t A) {
   return ((A * x - 5 * A) * x + 8 * A) * x - 4 * A;
 }
 
 template <typename scalar_t>
-static inline void get_cubic_upsample_coefficients(
+inline void get_cubic_upsample_coefficients(
     scalar_t coeffs[4],
     scalar_t t) {
   scalar_t A = -0.75;
@@ -418,7 +418,7 @@ static inline void get_cubic_upsample_coefficients(
 }
 
 template <typename scalar_t>
-static inline scalar_t cubic_interp1d(
+inline scalar_t cubic_interp1d(
     scalar_t x0,
     scalar_t x1,
     scalar_t x2,
@@ -434,7 +434,7 @@ static inline scalar_t cubic_interp1d(
 // type can accurately represent, the type casting to `int64_t` might exceed
 // `input_size`, causing overflow. So we guard it with `std::min` below.
 template<typename scalar_t, typename opmath_t>
-static inline void guard_index_and_lambda(const opmath_t& real_input_index, const int64_t& input_size, int64_t& input_index, scalar_t& lambda) {
+inline void guard_index_and_lambda(const opmath_t& real_input_index, const int64_t& input_size, int64_t& input_index, scalar_t& lambda) {
   input_index = std::min(static_cast<int64_t>(floorf(real_input_index)), input_size - 1);
   lambda = std::min(
       std::max(real_input_index - input_index, static_cast<opmath_t>(0)),
@@ -443,7 +443,7 @@ static inline void guard_index_and_lambda(const opmath_t& real_input_index, cons
 }
 
 template<typename scalar_t, typename opmath_t>
-static inline void compute_source_index_and_lambda(
+inline void compute_source_index_and_lambda(
     int64_t& input_index0,
     int64_t& input_index1,
     scalar_t& lambda0,

aten/src/ATen/native/cpu/Loops.h

@@ -82,7 +82,7 @@ dereference_vec(char* C10_RESTRICT data[], const typename traits::result_type& o
 
 template <typename func_t,
     typename std::enable_if<!std::is_void<typename function_traits<func_t>::result_type>::value>::type* = nullptr>
-static inline void
+inline void
 execute_op(char* C10_RESTRICT data[], const int64_t* strides, int64_t i, int64_t n, func_t&& op) {
   using traits = function_traits<func_t>;
   using result_type = typename traits::result_type;
@@ -97,7 +97,7 @@ execute_op(char* C10_RESTRICT data[], const int64_t* strides, int64_t i, int64_t
 
 template <typename func_t,
     typename std::enable_if<std::is_void<typename function_traits<func_t>::result_type>::value>::type* = nullptr>
-static inline void
+inline void
 execute_op(char* C10_RESTRICT data[], const int64_t* strides, int64_t i, int64_t n, func_t&& op) {
   using traits = function_traits<func_t>;
   for (; i < n; i++) {
@@ -111,7 +111,7 @@ execute_op(char* C10_RESTRICT data[], const int64_t* strides, int64_t i, int64_t
 // Basic loop operation (one output, N inputs). May be auto-vectorized
 // by the compiler. Supports inputs and outputs of different types.
 template <typename func_t>
-static inline void
+inline void
 basic_loop(char* C10_RESTRICT data[], const int64_t* strides_, int64_t i, int64_t n, func_t&& op) {
   using traits = function_traits<func_t>;
   constexpr int ntensors = traits::arity + 1;
@@ -166,7 +166,7 @@ void handle_tuple_outputs(char* C10_RESTRICT data[],
 // 2. Iterate over the members of the returned tuple, set the corresponding
 //    output tensor by the tuple member in `handle_tuple_outputs` function.
 template <typename func_t>
-static inline void
+inline void
 multiple_outputs_loop(char* C10_RESTRICT data[], const int64_t* strides_, int64_t i, int64_t n, func_t&& op) {
   using traits = function_traits<func_t>;
 
@@ -195,7 +195,7 @@ multiple_outputs_loop(char* C10_RESTRICT data[], const int64_t* strides_, int64_
 // a scalar (stride 0). It's position is indicated by the argument `S`. If `S`
 // is 0, then there are no scalar inputs.
 template <typename func_t, typename vec_func_t>
-static inline void
+inline void
 vectorized_loop(char** C10_RESTRICT data_, int64_t n, int64_t S, func_t&& op, vec_func_t&& vop) {
   using traits = function_traits<vec_func_t>;
   using scalar_t = typename function_traits<func_t>::result_type;
@@ -228,7 +228,7 @@ vectorized_loop(char** C10_RESTRICT data_, int64_t n, int64_t S, func_t&& op, ve
 
 
 template <typename traits, typename cb_t>
-static inline void unroll_contiguous_scalar_checks(
+inline void unroll_contiguous_scalar_checks(
     const int64_t* /*strides*/,
     std::index_sequence<>,
     cb_t&& cb) {
@@ -236,7 +236,7 @@ static inline void unroll_contiguous_scalar_checks(
 }
 
 template <typename traits, typename cb_t, size_t INDEX0, size_t ...INDEX>
-static inline void unroll_contiguous_scalar_checks(
+inline void unroll_contiguous_scalar_checks(
     const int64_t* strides,
     std::index_sequence<INDEX0, INDEX...>,
     cb_t&& cb) {

aten/src/ATen/native/cpu/Reduce.h

@@ -21,21 +21,21 @@ using namespace vec;
 
 // reduction that is contiguous over the input in dim 0
 template <typename traits>
-static inline bool is_contiguous_reduction(const int64_t* strides) {
+inline bool is_contiguous_reduction(const int64_t* strides) {
   return strides[0] == 0 &&
          strides[1] == sizeof(typename traits::arg2_t);
 }
 
 // reduction that is contiguous over the input in dim 1
 template <typename traits>
-static inline bool is_outer_reduction(const int64_t* strides) {
+inline bool is_outer_reduction(const int64_t* strides) {
   return strides[0] == 0 &&
          strides[2] == sizeof(typename traits::result_type) &&
          strides[3] == sizeof(typename traits::arg2_t);
 }
 
 template <typename func_t, typename vec_func_t>
-static inline void vectorized_reduction(char** data, int64_t n, int64_t stride,
+inline void vectorized_reduction(char** data, int64_t n, int64_t stride,
                                         func_t op, vec_func_t vop, bool reduce) {
   VEC_LOOP_HEADER(func_t, data)
   const char* in1_ptr = data[1];
@@ -69,7 +69,7 @@ static inline void vectorized_reduction(char** data, int64_t n, int64_t stride,
 }
 
 template <typename F>
-static inline void UNARY_OUTER_LOOP(char* data[2], const int64_t strides[2], int64_t n, F f) {
+inline void UNARY_OUTER_LOOP(char* data[2], const int64_t strides[2], int64_t n, F f) {
   for (const auto j C10_UNUSED : c10::irange(n)) {
     f();
     data[0] += strides[0];
@@ -79,7 +79,7 @@ static inline void UNARY_OUTER_LOOP(char* data[2], const int64_t strides[2], int
 
 // computes the reduction out = op(out, in)
 template <typename func_t, typename vec_func_t>
-static inline void vectorized_inner_reduction(char** data, int64_t n, func_t op, vec_func_t vop) {
+inline void vectorized_inner_reduction(char** data, int64_t n, func_t op, vec_func_t vop) {
   VEC_LOOP_HEADER(func_t, data)
   int64_t vector_stride = 4 * Vec::size() * sizeof(scalar_t);
   int64_t count = n / (4 * Vec::size());
@@ -93,7 +93,7 @@ static inline void vectorized_inner_reduction(char** data, int64_t n, func_t op,
 
 // computes the reduction out = op(out, in)
 template <typename func_t, typename vec_func_t>
-static inline void vectorized_outer_reduction(char** data, int64_t inner_stride, int64_t size0, int64_t size1, func_t op, vec_func_t vop) {
+inline void vectorized_outer_reduction(char** data, int64_t inner_stride, int64_t size0, int64_t size1, func_t op, vec_func_t vop) {
   VEC_LOOP_HEADER(func_t, data)
 
   // reduce down each column of 4 * Vec::size() elements (128 or 256 bytes)
@@ -132,13 +132,13 @@ static void set_results(const res_t result, const TensorIteratorBase &iter, cons
 }
 
 template<typename traits, std::size_t i = 0, typename... tuple_t>
-static inline typename std::enable_if<i == sizeof...(tuple_t), std::size_t>::type
+inline typename std::enable_if<i == sizeof...(tuple_t), std::size_t>::type
 for_each_in_tuple(const std::tuple<tuple_t...>& /*t*/, const TensorIteratorBase& /*iter*/, const int /*num_outputs*/) {
   return i;
 }
 
 template<typename traits, std::size_t i = 0, typename... tuple_t>
-static inline typename std::enable_if<i < sizeof...(tuple_t), std::size_t>::type
+inline typename std::enable_if<i < sizeof...(tuple_t), std::size_t>::type
 for_each_in_tuple(const std::tuple<tuple_t...>& t, const TensorIteratorBase &iter, const int num_outputs) {
   if (i < (size_t)num_outputs) {
     set_result<traits>(i, std::get<i>(t), iter, num_outputs);
@@ -286,7 +286,7 @@ void binary_kernel_reduce_vec(TensorIteratorBase& iter, func_t op, vec_func_t vo
 // when reduction is on most inner dimension (dim 0 in TensorIterator)
 // and input has contiguous most inner dimension, `binary_kernel_reduce_lastdim`
 // can be used.
-static inline bool is_reduce_lastdim(TensorIteratorBase& iter) {
+inline bool is_reduce_lastdim(TensorIteratorBase& iter) {
   return iter.num_reduce_dims() == 1 && iter.is_dim_reduced(0)
       && iter.ninputs() == 1 && iter.strides(1)[0] == iter.element_size(1);
 }