From 6eebf8e853569669c09b4b4fb89cefadbe18b0fd Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Thu, 28 Aug 2025 11:35:54 -0700
Subject: [PATCH 01/15] simplify the release linux workflow

---
 .github/scripts/generate-release-matrix.py    |  8 +--
 .github/workflows/build_wheels_linux.yml      | 55 ++++++++++++++++---
 ...ase-linux.yml => release-linux-x86_64.yml} | 15 ++---
 3 files changed, 59 insertions(+), 19 deletions(-)
 rename .github/workflows/{release-linux.yml => release-linux-x86_64.yml} (93%)

diff --git a/.github/scripts/generate-release-matrix.py b/.github/scripts/generate-release-matrix.py
index 4b232026de..7b9e0263dd 100644
--- a/.github/scripts/generate-release-matrix.py
+++ b/.github/scripts/generate-release-matrix.py
@@ -5,16 +5,16 @@
 import sys
 
 RELEASE_CUDA_VERSION = {
-    "wheel": ["cu128"],
-    "tarball": ["cu128"],
+    "wheel": ["cu129"],
+    "tarball": ["cu129"],
 }
 RELEASE_PYTHON_VERSION = {
-    "wheel": ["3.9", "3.10", "3.11", "3.12", "3.13"],
+    "wheel": ["3.10", "3.11", "3.12", "3.13"],
     "tarball": ["3.11"],
 }
 
 CXX11_TARBALL_CONTAINER_IMAGE = {
-    "cu128": "pytorch/libtorch-cxx11-builder:cuda12.8-main",
+    "cu129": "pytorch/libtorch-cxx11-builder:cuda12.9-main",
 }
 
 
diff --git a/.github/workflows/build_wheels_linux.yml b/.github/workflows/build_wheels_linux.yml
index 4b95a90e86..0e63881512 100644
--- a/.github/workflows/build_wheels_linux.yml
+++ b/.github/workflows/build_wheels_linux.yml
@@ -117,6 +117,16 @@ on:
         required: false
         default: false
         type: boolean
+      is-release-wheel:
+        description: Set to true if the build is for release wheel
+        required: false
+        default: false
+        type: boolean
+      is-release-tarball:
+        description: Set to true if the build is for release tarball
+        required: false
+        default: false
+        type: boolean
     secrets:
       PYPI_API_TOKEN:
         description: An optional token to upload to pypi
@@ -265,22 +275,41 @@ jobs:
           export PYTORCH_VERSION="$(${CONDA_RUN} pip show torch | grep ^Version: | sed 's/Version: *//' | sed 's/+.\+//')"
           ${CONDA_RUN} python setup.py clean
           echo "Successfully ran `python setup.py clean`"
-          if [[ "$BUILD_VERSION" != *"+"${CU_VERSION} ]]; then
-            BUILD_VERSION="${BUILD_VERSION}+${CU_VERSION}"
+          if [[ ${{ inputs.is-release-wheel }} == true || ${{ inputs.is-release-tarball }} == true ]]; then
+              # release version for upload to pypi
+              # BUILD_VERSION example: 2.4.0+cu121, we don't want the +cu121 part, so remove +cu121
+              BUILD_VERSION=${BUILD_VERSION%+*}
+              if [[ ${{ inputs.is-release-tarball }} == true ]]; then
+                  mkdir -p release/tarball
+                  TRT_VERSION=$(cat dev_dep_versions.yml | grep __tensorrt_version__ | sed 's/__tensorrt_version__: //g' | sed 's/"//g')
+                  bazel build //:libtorchtrt --compilation_mode opt --config=linux
+                  cp bazel-bin/libtorchtrt.tar.gz \
+                  release/tarball/libtorchtrt-${BUILD_VERSION}-tensorrt${TRT_VERSION}-cuda${CU_VERSION:2}-libtorch${PYTORCH_VERSION}-x86_64-linux.tar.gz
+              else
+                  if [[ ${{ inputs.is-release-wheel }} == true ]]; then
+                    param="--release"
+                  fi
+              fi
+          else
+              # release version for upload to pytorch index
+              if [[ "$BUILD_VERSION" != *"+"${CU_VERSION} ]]; then
+                BUILD_VERSION="${BUILD_VERSION}+${CU_VERSION}"
+              fi
           fi
+
           echo "BUILD_VERSION=$BUILD_VERSION"
           echo "USE_TRT_RTX=$USE_TRT_RTX"
           echo "LD_LIBRARY_PATH=$LD_LIBRARY_PATH"
           if [[ ${{ inputs.use-rtx }} == true ]]; then
             echo "Building tensorrt-rtx wheel"
-            ${CONDA_RUN} python setup.py bdist_wheel --use-rtx
+            ${CONDA_RUN} python setup.py bdist_wheel ${param} --use-rtx
           else
             if [[ ${{ inputs.is-jetpack }} == true ]]; then
               echo "Building tensorrt wheel for jetpack"
-              ${CONDA_RUN} python setup.py bdist_wheel --jetpack
+              ${CONDA_RUN} python setup.py bdist_wheel ${param} --jetpack
             else
               echo "Building standard tensorrt wheel"
-              ${CONDA_RUN} python setup.py bdist_wheel
+              ${CONDA_RUN} python setup.py bdist_wheel ${param}
             fi
           fi
       - name: Repair Manylinux_2_28 Wheel
@@ -289,7 +318,7 @@ jobs:
           PACKAGE_NAME: ${{ inputs.package-name }}
           SMOKE_TEST_SCRIPT: ${{ inputs.smoke-test-script }}
         # TODO: lan to verify whether manylinux repair is needed for jetpack
-        #if: ${{ inputs.is-jetpack == true }}
+        if: ${{ inputs.is-release-tarball == false }}
         run: |
           set -euxo pipefail
           source "${BUILD_ENV_FILE}"
@@ -312,6 +341,7 @@ jobs:
         env:
           PACKAGE_NAME: ${{ inputs.package-name }}
           SMOKE_TEST_SCRIPT: ${{ inputs.smoke-test-script }}
+        if: ${{ inputs.is-release-tarball == false }}
         run: |
           set -euxo pipefail
           source "${BUILD_ENV_FILE}"
@@ -348,18 +378,27 @@ jobs:
       # NB: Only upload to GitHub after passing smoke tests
 
       - name: Upload wheel to GitHub
+        if: ${{ inputs.is-release-wheel == false }}
         continue-on-error: true
         uses: actions/upload-artifact@v4
         with:
           name: ${{ env.ARTIFACT_NAME }}
           path: ${{ inputs.repository }}/dist/
+      - name: Upload tarball to GitHub
+        if: ${{ inputs.is-release-tarball == true }}
+        continue-on-error: true
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.ARTIFACT_NAME }}
+          path: ${{ inputs.repository }}/release/tarball/
 
   upload:
     needs: build
     name: upload-wheel-${{ matrix.python_version }}-${{ matrix.desired_cuda }}-${{ matrix.gpu_arch_type }}-${{ inputs.is-jetpack }}
     uses: pytorch/test-infra/.github/workflows/_binary_upload.yml@main
     #  if it is not jetpack nor rtx, then upload to pytorch index
-    if: ${{ inputs.is-jetpack == false && inputs.use-rtx == false }}
+    # if it is the release wheel or tarball, then skip upload to pytorch index
+    if: ${{ inputs.is-jetpack == false && inputs.use-rtx == false && inputs.is-release-wheel == false && inputs.is-release-tarball == false }}
     with:
       repository: ${{ inputs.repository }}
       ref: ${{ inputs.ref }}
@@ -373,5 +412,5 @@ jobs:
       PYPI_API_TOKEN: ${{ secrets.PYPI_API_TOKEN }}
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{inputs.use-rtx}}-${{inputs.architecture}}-${{inputs.is-jetpack}}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{inputs.is-release-wheel}}-${{inputs.is-release-tarball}}-${{inputs.use-rtx}}-${{inputs.architecture}}-${{inputs.is-jetpack}}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}
   cancel-in-progress: true
diff --git a/.github/workflows/release-linux.yml b/.github/workflows/release-linux-x86_64.yml
similarity index 93%
rename from .github/workflows/release-linux.yml
rename to .github/workflows/release-linux-x86_64.yml
index 2c4be0ec1d..da5e1bd06a 100644
--- a/.github/workflows/release-linux.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -1,6 +1,7 @@
-name: Release Linux wheels and tarball artifacts
+name: Release x86_64 Linux wheels and tarball artifacts
 
 on:
+  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
@@ -59,8 +60,8 @@ jobs:
             env-var-script: packaging/env_vars.txt
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
-            cxx11-tarball-release: "true"
-    uses: ./.github/workflows/release-wheel-linux.yml
+            is-release-tarball: "true"
+    uses: ./.github/workflows/build_wheels_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
@@ -73,7 +74,7 @@ jobs:
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-      cxx11-tarball-release: "true"
+      is-release-tarball: "true"
 
   generate-release-wheel-matrix:
     needs: [generate-matrix]
@@ -109,8 +110,8 @@ jobs:
             env-var-script: packaging/env_vars.txt
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
-            cxx11-tarball-release: "true"
-    uses: ./.github/workflows/release-wheel-linux.yml
+            is-release-wheel: "true"
+    uses: ./.github/workflows/build_wheels_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
@@ -123,7 +124,7 @@ jobs:
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-      cxx11-tarball-release: "false"
+      is-release-wheel: "true"
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}

From 2fb253a69867b39613804335e47403a7e6d30df8 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Thu, 28 Aug 2025 13:00:09 -0700
Subject: [PATCH 02/15] test

---
 .github/workflows/build_wheels_linux.yml    |   3 +-
 .github/workflows/build_wheels_windows.yml  |  34 ++-
 .github/workflows/release-linux-aarch64.yml | 131 ++++++++++
 .github/workflows/release-linux-x86_64.yml  |   6 +-
 .github/workflows/release-wheel-linux.yml   | 249 --------------------
 .github/workflows/release-wheel-windows.yml | 213 -----------------
 .github/workflows/release-windows.yml       |  81 ++++++-
 7 files changed, 237 insertions(+), 480 deletions(-)
 create mode 100644 .github/workflows/release-linux-aarch64.yml
 delete mode 100644 .github/workflows/release-wheel-linux.yml
 delete mode 100644 .github/workflows/release-wheel-windows.yml

diff --git a/.github/workflows/build_wheels_linux.yml b/.github/workflows/build_wheels_linux.yml
index 0e63881512..bdcff46256 100644
--- a/.github/workflows/build_wheels_linux.yml
+++ b/.github/workflows/build_wheels_linux.yml
@@ -285,6 +285,7 @@ jobs:
                   bazel build //:libtorchtrt --compilation_mode opt --config=linux
                   cp bazel-bin/libtorchtrt.tar.gz \
                   release/tarball/libtorchtrt-${BUILD_VERSION}-tensorrt${TRT_VERSION}-cuda${CU_VERSION:2}-libtorch${PYTORCH_VERSION}-x86_64-linux.tar.gz
+                  return 0
               else
                   if [[ ${{ inputs.is-release-wheel }} == true ]]; then
                     param="--release"
@@ -378,7 +379,7 @@ jobs:
       # NB: Only upload to GitHub after passing smoke tests
 
       - name: Upload wheel to GitHub
-        if: ${{ inputs.is-release-wheel == false }}
+        if: ${{ inputs.is-release-tarball == false }}
         continue-on-error: true
         uses: actions/upload-artifact@v4
         with:
diff --git a/.github/workflows/build_wheels_windows.yml b/.github/workflows/build_wheels_windows.yml
index 5c9e391cd6..8eab386033 100644
--- a/.github/workflows/build_wheels_windows.yml
+++ b/.github/workflows/build_wheels_windows.yml
@@ -89,6 +89,16 @@ on:
         description: 'CPU architecture to build for'
         default: "x64"
         type: string
+      is-release-tarball:
+        description: Set to true if the build is for release tarball
+        required: false
+        default: false
+        type: boolean
+      is-release-wheel:
+        description: Set to true if the build is for release wheel
+        required: false
+        default: false
+        type: boolean
 
 permissions:
   id-token: write
@@ -277,6 +287,19 @@ jobs:
           BUILD_PARAMS: ${{ inputs.wheel-build-params }}
         run: |
           source "${BUILD_ENV_FILE}"
+          if [[ ${{ inputs.is-release-wheel }} == true || ${{ inputs.is-release-tarball }} == true ]]; then
+            # release version for upload to pypi
+            # BUILD_VERSION example: 2.4.0+cu121, we don't want the +cu121 part, so remove +cu121
+            BUILD_VERSION=${BUILD_VERSION%+*}
+            if [[ ${{ inputs.is-release-tarball }} == true ]]; then
+              mkdir -p release/tarball
+              TRT_VERSION=$(cat dev_dep_versions.yml | grep __tensorrt_version__ | sed 's/__tensorrt_version__: //g' | sed 's/"//g')
+              bazel build //:libtorchtrt --compilation_mode opt --config=windows
+              cp bazel-bin/libtorchtrt.tar.gz \
+              release/tarball/libtorchtrt-${BUILD_VERSION}-tensorrt${TRT_VERSION}-cuda${CU_VERSION:2}-libtorch${PYTORCH_VERSION}-x86_64-windows.tar.gz
+              return 0
+            fi
+          fi
           if [[ -z "${ENV_SCRIPT}" ]]; then
             ${CONDA_RUN} python setup.py bdist_wheel
           else
@@ -365,11 +388,19 @@ jobs:
         run: |
           echo "ARTIFACT_NAME=${REPOSITORY//\//_}_${REF//\//_}_${PYTHON_VERSION}_${CU_VERSION}_${ARCH}" >> "${GITHUB_ENV}"
       - name: Upload wheel to GitHub
+        if: ${{ inputs.is-release-tarball == false }}
         continue-on-error: true
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: ${{ env.ARTIFACT_NAME }}
           path: ${{ inputs.repository }}/dist/
+      - name: Upload tarball to GitHub
+        if: ${{ inputs.is-release-tarball == true }}
+        continue-on-error: true
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.ARTIFACT_NAME }}
+          path: ${{ inputs.repository }}/release/tarball/
       - uses: ./test-infra/.github/actions/teardown-windows
         if: inputs.architecture == 'x64'
         name: Teardown Windows
@@ -378,7 +409,8 @@ jobs:
     needs: build
     uses: pytorch/test-infra/.github/workflows/_binary_upload.yml@main
     # for tensorrt-rtx build, do not upload to pytorch indexat at all
-    if: ${{ inputs.use-rtx == false }}
+    # if it is the release wheel or tarball, then skip upload to pytorch index(release wheel are uploaded to pypi)
+    if: ${{ inputs.use-rtx == false && inputs.is-release-wheel == false && inputs.is-release-tarball == false }}
     with:
       repository: ${{ inputs.repository }}
       ref: ${{ inputs.ref }}
diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
new file mode 100644
index 0000000000..85ab3672a9
--- /dev/null
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -0,0 +1,131 @@
+name: Release aarch64 Linux wheels and tarball artifacts
+
+on:
+  pull_request:
+  push:
+    tags:
+      # NOTE: Binary build pipelines should only get triggered on release candidate builds
+      # Release candidate tags look like: v1.11.0-rc1
+      - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
+  workflow_dispatch:
+
+permissions:
+  id-token: write
+  contents: read
+  packages: write
+
+jobs:
+  generate-matrix:
+    uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
+    if: ${{ contains(github.event.pull_request.labels.*.name, 'build-release-artifacts') || startsWith(github.event.ref, 'refs/tags/v') }}
+    with:
+      package-type: wheel
+      os: linux
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      with-rocm: false
+      with-cpu: false
+
+  generate-release-tarball-matrix:
+    needs: [generate-matrix]
+    outputs:
+      matrix: ${{ steps.generate.outputs.matrix }}
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - uses: actions/checkout@v4
+        with:
+          repository: pytorch/tensorrt
+      - name: Generate release matrix
+        id: generate
+        run: |
+          set -eou pipefail
+          MATRIX_BLOB=${{ toJSON(needs.generate-matrix.outputs.matrix) }}
+          MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --tarball_matrix "${MATRIX_BLOB}")"
+          echo "${MATRIX_BLOB}"
+          echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
+
+  release-cxx11-tarball-artifacts:
+    needs: [generate-release-tarball-matrix]
+    name: Release torch-tensorrt cxx11 tarball artifacts
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - repository: pytorch/tensorrt
+            package-name: torch_tensorrt
+            pre-script: packaging/pre_build_script.sh
+            env-var-script: packaging/env_vars.txt
+            post-script: packaging/post_build_script.sh
+            smoke-test-script: packaging/smoke_test_script.sh
+            is-release-tarball: "true"
+    uses: ./.github/workflows/build_wheels_linux.yml
+    with:
+      repository: ${{ matrix.repository }}
+      ref: ""
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      build-matrix: ${{ needs.generate-release-tarball-matrix.outputs.matrix }}
+      pre-script: ${{ matrix.pre-script }}
+      env-var-script: ${{ matrix.env-var-script }}
+      post-script: ${{ matrix.post-script }}
+      package-name: ${{ matrix.package-name }}
+      smoke-test-script: ${{ matrix.smoke-test-script }}
+      trigger-event: ${{ github.event_name }}
+      is-release-tarball: true
+
+  generate-release-wheel-matrix:
+    needs: [generate-matrix]
+    outputs:
+      matrix: ${{ steps.generate.outputs.matrix }}
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - uses: actions/checkout@v4
+        with:
+          repository: pytorch/tensorrt
+      - name: Generate release matrix
+        id: generate
+        run: |
+          set -eou pipefail
+          MATRIX_BLOB=${{ toJSON(needs.generate-matrix.outputs.matrix) }}
+          MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --wheel_matrix "${MATRIX_BLOB}")"
+          echo "${MATRIX_BLOB}"
+          echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
+
+  release-wheel-artifacts:
+    name: Release torch-tensorrt wheel artifacts
+    needs: [generate-release-wheel-matrix]
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - repository: pytorch/tensorrt
+            package-name: torch_tensorrt
+            pre-script: packaging/pre_build_script.sh
+            env-var-script: packaging/env_vars.txt
+            post-script: packaging/post_build_script.sh
+            smoke-test-script: packaging/smoke_test_script.sh
+            is-release-wheel:  true
+    uses: ./.github/workflows/build_wheels_linux.yml
+    with:
+      repository: ${{ matrix.repository }}
+      ref: ""
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      build-matrix: ${{ needs.generate-release-wheel-matrix.outputs.matrix }}
+      pre-script: ${{ matrix.pre-script }}
+      env-var-script: ${{ matrix.env-var-script }}
+      post-script: ${{ matrix.post-script }}
+      package-name: ${{ matrix.package-name }}
+      smoke-test-script: ${{ matrix.smoke-test-script }}
+      trigger-event: ${{ github.event_name }}
+      is-release-wheel: true
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
+  cancel-in-progress: true
\ No newline at end of file
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index da5e1bd06a..e0190c2a72 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -74,7 +74,7 @@ jobs:
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-      is-release-tarball: "true"
+      is-release-tarball: true
 
   generate-release-wheel-matrix:
     needs: [generate-matrix]
@@ -110,7 +110,7 @@ jobs:
             env-var-script: packaging/env_vars.txt
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
-            is-release-wheel: "true"
+            is-release-wheel:  true
     uses: ./.github/workflows/build_wheels_linux.yml
     with:
       repository: ${{ matrix.repository }}
@@ -124,7 +124,7 @@ jobs:
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-      is-release-wheel: "true"
+      is-release-wheel: true
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
diff --git a/.github/workflows/release-wheel-linux.yml b/.github/workflows/release-wheel-linux.yml
deleted file mode 100644
index efd73f490f..0000000000
--- a/.github/workflows/release-wheel-linux.yml
+++ /dev/null
@@ -1,249 +0,0 @@
-name: Build Release Wheel Artifacts on Linux
-
-on:
-  workflow_call:
-    inputs:
-      repository:
-        description: 'Repository to checkout, defaults to ""'
-        default: ""
-        type: string
-      ref:
-        description: 'Reference to checkout, defaults to "nightly"'
-        default: "nightly"
-        type: string
-      test-infra-repository:
-        description: "Test infra repository to use"
-        default: "pytorch/test-infra"
-        type: string
-      test-infra-ref:
-        description: "Test infra reference to use"
-        default: ""
-        type: string
-      build-matrix:
-        description: "Build matrix to utilize"
-        default: ""
-        type: string
-      pre-script:
-        description: "Pre script to run prior to build"
-        default: ""
-        type: string
-      post-script:
-        description: "Post script to run prior to build"
-        default: ""
-        type: string
-      smoke-test-script:
-        description: "Script for Smoke Test for a specific domain"
-        default: ""
-        type: string
-      env-var-script:
-        description: "Script that sets Domain-Specific Environment Variables"
-        default: ""
-        type: string
-      package-name:
-        description: "Name of the actual python package that is imported"
-        default: ""
-        type: string
-      trigger-event:
-        description: "Trigger Event in caller that determines whether or not to upload"
-        default: ""
-        type: string
-      cache-path:
-        description: "The path(s) on the runner to cache or restore. The path is relative to repository."
-        default: ""
-        type: string
-      cache-key:
-        description: "The key created when saving a cache and the key used to search for a cache."
-        default: ""
-        type: string
-      architecture:
-        description: Architecture to build for x86_64 for default Linux, or aarch64 for Linux aarch64 builds
-        required: false
-        type: string
-        default: x86_64
-      submodules:
-        description: Works as stated in actions/checkout, but the default value is recursive
-        required: false
-        type: string
-        default: recursive
-      setup-miniconda:
-        description: Set to true if setup-miniconda is needed
-        required: false
-        type: boolean
-        default: true
-      cxx11-tarball-release:
-        description: "Flag whether this is the cxx11 tarball release"
-        required: true
-        type: string
-        default: "false"
-
-permissions:
-  id-token: write
-  contents: read
-
-jobs:
-  release:
-    strategy:
-      fail-fast: false
-      matrix: ${{ fromJSON(inputs.build-matrix) }}
-    env:
-      PYTHON_VERSION: ${{ matrix.python_version }}
-      PACKAGE_TYPE: wheel
-      REPOSITORY: ${{ inputs.repository }}
-      REF: ${{ inputs.ref }}
-      CU_VERSION: ${{ matrix.desired_cuda }}
-      UPLOAD_TO_BASE_BUCKET: ${{ matrix.upload_to_base_bucket }}
-      ARCH: ${{ inputs.architecture }}
-    name: release_${{ matrix.build_name }}
-    runs-on: ${{ matrix.validation_runner }}
-    container:
-      image: ${{ matrix.container_image }}
-      options: ${{ matrix.gpu_arch_type == 'cuda' && '--gpus all' || ' ' }}
-    # If a build is taking longer than 120 minutes on these runners we need
-    # to have a conversation
-    timeout-minutes: 120
-
-    steps:
-      - name: Clean workspace
-        shell: bash -l {0}
-        run: |
-          set -x
-          echo "::group::Cleanup debug output"
-          rm -rf "${GITHUB_WORKSPACE}"
-          mkdir -p "${GITHUB_WORKSPACE}"
-          if [[ "${{ inputs.architecture }}" = "aarch64" ]]; then
-            rm -rf "${RUNNER_TEMP}/*"
-          fi
-          echo "::endgroup::"
-      - uses: actions/checkout@v4
-        with:
-          # Support the use case where we need to checkout someone's fork
-          repository: ${{ inputs.test-infra-repository }}
-          ref: ${{ inputs.test-infra-ref }}
-          path: test-infra
-      - uses: actions/checkout@v4
-        if: ${{ env.ARCH == 'aarch64' }}
-        with:
-          # Support the use case where we need to checkout someone's fork
-          repository: "pytorch/builder"
-          ref: "main"
-          path: builder
-      - name: Set linux aarch64 CI
-        if: ${{ inputs.architecture == 'aarch64' }}
-        shell: bash -l {0}
-        env:
-          DESIRED_PYTHON: ${{ matrix.python_version }}
-        run: |
-          set +e
-          # TODO: This is temporary aarch64 setup script, this should be integrated into aarch64 docker.
-          ${GITHUB_WORKSPACE}/builder/aarch64_linux/aarch64_ci_setup.sh
-          echo "/opt/conda/bin" >> $GITHUB_PATH
-          set -e
-      - uses: ./test-infra/.github/actions/set-channel
-      - name: Set PYTORCH_VERSION
-        if: ${{ env.CHANNEL == 'test' }}
-        run: |
-          # When building RC, set the version to be the current candidate version,
-          # otherwise, leave it alone so nightly will pick up the latest
-          echo "PYTORCH_VERSION=${{ matrix.stable_version }}" >> "${GITHUB_ENV}"
-      - uses: ./test-infra/.github/actions/setup-binary-builds
-        env:
-          PLATFORM: ${{ inputs.architecture == 'aarch64'  && 'linux-aarch64' || ''}}
-        with:
-          repository: ${{ inputs.repository }}
-          ref: ${{ inputs.ref }}
-          submodules: ${{ inputs.submodules }}
-          setup-miniconda: ${{ inputs.setup-miniconda }}
-          python-version: ${{ env.PYTHON_VERSION }}
-          cuda-version: ${{ env.CU_VERSION }}
-          arch: ${{ env.ARCH }}
-      - name: Combine Env Var and Build Env Files
-        if: ${{ inputs.env-var-script != '' }}
-        working-directory: ${{ inputs.repository }}
-        shell: bash -l {0}
-        run: |
-          cat "${{ inputs.env-var-script }}" >> "${BUILD_ENV_FILE}"
-      - name: Install torch dependency
-        shell: bash -l {0}
-        run: |
-          set -x
-          # shellcheck disable=SC1090
-          source "${BUILD_ENV_FILE}"
-          # shellcheck disable=SC2086
-          ${CONDA_RUN} ${PIP_INSTALL_TORCH}
-      - name: Run Pre-Script with Caching
-        if: ${{ inputs.pre-script != '' }}
-        uses: ./test-infra/.github/actions/run-script-with-cache
-        with:
-          cache-path: ${{ inputs.cache-path }}
-          cache-key: ${{ inputs.cache-key }}
-          repository: ${{ inputs.repository  }}
-          script: ${{ inputs.pre-script }}
-      - name: Build clean
-        working-directory: ${{ inputs.repository }}
-        shell: bash -l {0}
-        run: |
-          set -x
-          source "${BUILD_ENV_FILE}"
-          ${CONDA_RUN} python setup.py clean
-      - name: Build the wheel (bdist_wheel)
-        working-directory: ${{ inputs.repository }}
-        shell: bash -l {0}
-        run: |
-          set -x
-          source "${BUILD_ENV_FILE}"
-
-          # BUILD_VERSION example: 2.4.0+cu121, we don't want the +cu121 part, so remove +cu121
-          BUILD_VERSION=${BUILD_VERSION%+*}
-          TRT_VERSION=$(cat dev_dep_versions.yml | grep __tensorrt_version__ | sed 's/__tensorrt_version__: //g' | sed 's/"//g')
-          mkdir release
-          mkdir release/tarball
-          mkdir release/wheel
-          if [[ "${{ inputs.cxx11-tarball-release }}" == "true" ]]; then
-              bazel build //:libtorchtrt --compilation_mode opt --config=linux
-              cp bazel-bin/libtorchtrt.tar.gz \
-              release/tarball/libtorchtrt-${BUILD_VERSION}-tensorrt${TRT_VERSION}-cuda${CU_VERSION:2}-libtorch${PYTORCH_VERSION}-x86_64-linux.tar.gz
-          else
-              ${CONDA_RUN} python setup.py bdist_wheel --release
-              ${CONDA_RUN} python -m pip install auditwheel
-              ${CONDA_RUN} python -m auditwheel repair \
-              $(cat py/ci/soname_excludes.params) \
-                --plat manylinux_2_34_x86_64 \
-                dist/torch_tensorrt-*-linux_x86_64.whl
-              ${CONDA_RUN} python -m zipfile --list wheelhouse/torch_tensorrt-*_x86_64.whl
-              cp wheelhouse/torch_tensorrt-*_x86_64.whl release/wheel/
-          fi
-      - name: Run Post-Script
-        if: ${{ inputs.post-script != '' }}
-        uses: ./test-infra/.github/actions/run-script-with-cache
-        with:
-          repository: ${{ inputs.repository  }}
-          script: ${{ inputs.post-script }}
-      - name: Smoke Test
-        shell: bash -l {0}
-        env:
-          PACKAGE_NAME: ${{ inputs.package-name }}
-          SMOKE_TEST_SCRIPT: ${{ inputs.smoke-test-script }}
-        run: |
-          set -x
-          source "${BUILD_ENV_FILE}"
-          # TODO: add smoke test for the auditwheel tarball built
-
-      # NB: Only upload to GitHub after passing smoke tests
-      - name: Upload wheel to GitHub
-        if: ${{ inputs.cxx11-tarball-release != 'true' }}
-        continue-on-error: true
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.ARTIFACT_NAME }}
-          path: ${{ inputs.repository }}/release/wheel/
-      - name: Upload cxx11 tarball to GitHub
-        if: ${{ inputs.cxx11-tarball-release == 'true' }}
-        continue-on-error: true
-        uses: actions/upload-artifact@v4
-        with:
-          name: cxx11-tarball-${{ env.PYTHON_VERSION }}-${{ env.CU_VERSION }}
-          path: ${{ inputs.repository }}/release/tarball/
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}-${{ inputs.cxx11-tarball-release }}
-  cancel-in-progress: true
\ No newline at end of file
diff --git a/.github/workflows/release-wheel-windows.yml b/.github/workflows/release-wheel-windows.yml
deleted file mode 100644
index 2ea88bce9e..0000000000
--- a/.github/workflows/release-wheel-windows.yml
+++ /dev/null
@@ -1,213 +0,0 @@
-name: Build Release Wheel Artifacts on Windows
-
-on:
-  workflow_call:
-    inputs:
-      repository:
-        description: 'Repository to checkout, defaults to ""'
-        default: ""
-        type: string
-      ref:
-        description: 'Reference to checkout, defaults to "nightly"'
-        default: "nightly"
-        type: string
-      test-infra-repository:
-        description: "Test infra repository to use"
-        default: "pytorch/test-infra"
-        type: string
-      test-infra-ref:
-        description: "Test infra reference to use"
-        default: ""
-        type: string
-      build-matrix:
-        description: "Build matrix to utilize"
-        default: ""
-        type: string
-      pre-script:
-        description: "Pre script to run prior to build"
-        default: ""
-        type: string
-      env-script:
-        description: "Script to setup environment variables for the build"
-        default: ""
-        type: string
-      wheel-build-params:
-        description: "Additional parameters for bdist_wheel"
-        default: ""
-        type: string
-      post-script:
-        description: "Post script to run prior to build"
-        default: ""
-        type: string
-      smoke-test-script:
-        description: "Script for Smoke Test for a specific domain"
-        default: ""
-        type: string
-      package-name:
-        description: "Name of the actual python package that is imported"
-        default: ""
-        type: string
-      trigger-event:
-        description: "Trigger Event in caller that determines whether or not to upload"
-        default: ""
-        type: string
-      cache-path:
-        description: "The path(s) on the runner to cache or restore. The path is relative to repository."
-        default: ""
-        type: string
-      cache-key:
-        description: "The key created when saving a cache and the key used to search for a cache."
-        default: ""
-        type: string
-      submodules:
-        description: "Works as stated in actions/checkout, but the default value is recursive"
-        required: false
-        type: string
-        default: recursive
-
-permissions:
-  id-token: write
-  contents: read
-
-jobs:
-  build-release-wheel:
-    strategy:
-      fail-fast: false
-      matrix: ${{ fromJSON(inputs.build-matrix) }}
-    env:
-      PYTHON_VERSION: ${{ matrix.python_version }}
-      PACKAGE_TYPE: wheel
-      REPOSITORY: ${{ inputs.repository }}
-      REF: ${{ inputs.ref }}
-      CU_VERSION: ${{ matrix.desired_cuda }}
-      UPLOAD_TO_BASE_BUCKET: ${{ matrix.upload_to_base_bucket }}
-    name: ${{ matrix.build_name }}
-    runs-on: ${{ matrix.validation_runner }}
-    defaults:
-      run:
-        shell: bash -l {0}
-    # If a build is taking longer than 120 minutes on these runners we need
-    # to have a conversation
-    timeout-minutes: 120
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          # Support the use case where we need to checkout someone's fork
-          repository: ${{ inputs.test-infra-repository }}
-          ref: ${{ inputs.test-infra-ref }}
-          path: test-infra
-      - uses: ./test-infra/.github/actions/setup-ssh
-        name: Setup SSH
-        with:
-          github-secret: ${{ secrets.GITHUB_TOKEN }}
-          activate-with-label: false
-          instructions: "SSH with rdesktop using ssh -L 3389:localhost:3389 %%username%%@%%hostname%%"
-      - name: Add Conda scripts to GitHub path
-        run: |
-          echo "C:/Jenkins/Miniconda3/Scripts" >> $GITHUB_PATH
-      - uses: ./test-infra/.github/actions/set-channel
-      - name: Set PYTORCH_VERSION
-        if: ${{ env.CHANNEL == 'test' }}
-        run: |
-          # When building RC, set the version to be the current candidate version,
-          # otherwise, leave it alone so nightly will pick up the latest
-          echo "PYTORCH_VERSION=${{ matrix.stable_version }}" >> "${GITHUB_ENV}"
-      - uses: ./test-infra/.github/actions/setup-binary-builds
-        with:
-          repository: ${{ inputs.repository }}
-          ref: ${{ inputs.ref }}
-          submodules: ${{ inputs.submodules }}
-          setup-miniconda: false
-          python-version: ${{ env.PYTHON_VERSION }}
-          cuda-version: ${{ env.CU_VERSION }}
-          arch: ${{ env.ARCH }}
-      - name: Install torch dependency
-        run: |
-          source "${BUILD_ENV_FILE}"
-          # shellcheck disable=SC2086
-          ${CONDA_RUN} ${PIP_INSTALL_TORCH}
-      - name: Run Pre-Script with Caching
-        if: ${{ inputs.pre-script != '' }}
-        uses: ./test-infra/.github/actions/run-script-with-cache
-        with:
-          cache-path: ${{ inputs.cache-path }}
-          cache-key: ${{ inputs.cache-key }}
-          repository: ${{ inputs.repository  }}
-          script: ${{ inputs.pre-script }}
-          is_windows: 'enabled'
-      - name: Build clean
-        working-directory: ${{ inputs.repository }}
-        run: |
-          source "${BUILD_ENV_FILE}"
-          ${CONDA_RUN} python setup.py clean
-      - name: Build the wheel (bdist_wheel)
-        working-directory: ${{ inputs.repository }}
-        env:
-          ENV_SCRIPT: ${{ inputs.env-script }}
-          BUILD_PARAMS: ${{ inputs.wheel-build-params }}
-        run: |
-          set -x
-          source "${BUILD_ENV_FILE}"
-
-          if [[ "$CU_VERSION" == "cpu" ]]; then
-          # CUDA and CPU are ABI compatible on the CPU-only parts, so strip
-          # in this case
-            export PYTORCH_VERSION="$(${CONDA_RUN} pip show torch | grep ^Version: | sed 's/Version:  *//' | sed 's/+.\+//')"
-          else
-            export PYTORCH_VERSION="$(${CONDA_RUN} pip show torch | grep ^Version: | sed 's/Version:  *//')"
-          fi
-
-          # BUILD_VERSION example: 2.4.0+cu121, we don't want the +cu121 part, so remove +cu121
-          BUILD_VERSION=${BUILD_VERSION%+*}
-
-          if [[ -z "${ENV_SCRIPT}" ]]; then
-            ${CONDA_RUN} python setup.py bdist_wheel
-          else
-            if [[ ! -f ${ENV_SCRIPT} ]]; then
-              echo "::error::Specified env-script file (${ENV_SCRIPT}) not found"
-              exit 1
-            else
-              ${CONDA_RUN} ${ENV_SCRIPT} python setup.py bdist_wheel ${BUILD_PARAMS}
-            fi
-          fi
-      - name: Run post-script
-        working-directory: ${{ inputs.repository }}
-        env:
-          POST_SCRIPT: ${{ inputs.post-script }}
-          ENV_SCRIPT: ${{ inputs.env-script }}
-        if: ${{ inputs.post-script != '' }}
-        run: |
-          set -euxo pipefail
-          source "${BUILD_ENV_FILE}"
-          ${CONDA_RUN} ${ENV_SCRIPT} ${POST_SCRIPT}
-      - name: Smoke Test
-        env:
-          PACKAGE_NAME: ${{ inputs.package-name }}
-          SMOKE_TEST_SCRIPT: ${{ inputs.smoke-test-script }}
-        run: |
-          source "${BUILD_ENV_FILE}"
-          WHEEL_NAME=$(ls "${{ inputs.repository }}/dist/")
-          echo "$WHEEL_NAME"
-          ${CONDA_RUN} pip install "${{ inputs.repository }}/dist/$WHEEL_NAME"
-          if [[ ! -f "${{ inputs.repository }}"/${SMOKE_TEST_SCRIPT} ]]; then
-            echo "${{ inputs.repository }}/${SMOKE_TEST_SCRIPT} not found"
-            ${CONDA_RUN} python -c "import ${PACKAGE_NAME}; print('package version is ', ${PACKAGE_NAME}.__version__)"
-          else
-            echo "${{ inputs.repository }}/${SMOKE_TEST_SCRIPT} found"
-            ${CONDA_RUN} python "${{ inputs.repository }}/${SMOKE_TEST_SCRIPT}"
-          fi
-      # NB: Only upload to GitHub after passing smoke tests
-      - name: Upload wheel to GitHub
-        continue-on-error: true
-        uses: actions/upload-artifact@v4
-        with:
-          name: ${{ env.ARTIFACT_NAME }}
-          path: ${{ inputs.repository }}/dist/
-      - uses: ./test-infra/.github/actions/teardown-windows
-        if: always()
-        name: Teardown Windows
-
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
-  cancel-in-progress: true
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index 1a20eb7a71..bee7fc8721 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -1,6 +1,7 @@
 name: Release Windows wheels artifacts
 
 on:
+  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
@@ -25,8 +26,20 @@ jobs:
       with-rocm: false
       with-cpu: false
 
-  generate-release-matrix:
-    needs: [generate-matrix]
+  substitute-runner:
+    needs: generate-matrix
+    outputs:
+      matrix: ${{ steps.substitute.outputs.matrix }}
+    runs-on: ubuntu-latest
+    steps:
+      - name: Substitute runner
+        id: substitute
+        run: |
+          echo matrix="$(echo '${{ needs.generate-release-matrix.outputs.matrix }}' | sed -e 's/windows.g4dn.xlarge/windows.g5.4xlarge.nvidia.gpu/g')" >> ${GITHUB_OUTPUT}
+
+
+  generate-release-tarball-matrix:
+    needs: [substitute-runner]
     outputs:
       matrix: ${{ steps.generate.outputs.matrix }}
     runs-on: ubuntu-latest
@@ -37,28 +50,68 @@ jobs:
       - uses: actions/checkout@v4
         with:
           repository: pytorch/tensorrt
-      - name: Generate release matrix
+      - name: Generate release tarball matrix
         id: generate
         run: |
           set -eou pipefail
-          MATRIX_BLOB=${{ toJSON(needs.generate-matrix.outputs.matrix) }}
-          MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --wheel_matrix "${MATRIX_BLOB}")"
+          MATRIX_BLOB=${{ toJSON(needs.substitute-runner.outputs.matrix) }}
+          MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --tarball_matrix "${MATRIX_BLOB}")"
           echo "${MATRIX_BLOB}"
           echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
 
-  substitute-runner:
-    needs: generate-release-matrix
+  generate-release-wheel-matrix:
+    needs: [substitute-runner]
     outputs:
-      matrix: ${{ steps.substitute.outputs.matrix }}
+      matrix: ${{ steps.generate.outputs.matrix }}
     runs-on: ubuntu-latest
     steps:
-      - name: Substitute runner
-        id: substitute
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+      - uses: actions/checkout@v4
+        with:
+          repository: pytorch/tensorrt
+      - name: Generate release wheel matrix
+        id: generate
         run: |
-          echo matrix="$(echo '${{ needs.generate-release-matrix.outputs.matrix }}' | sed -e 's/windows.g4dn.xlarge/windows.g5.4xlarge.nvidia.gpu/g')" >> ${GITHUB_OUTPUT}
+          set -eou pipefail
+          MATRIX_BLOB=${{ toJSON(needs.substitute-runner.outputs.matrix) }}
+          MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --wheel_matrix "${MATRIX_BLOB}")"
+          echo "${MATRIX_BLOB}"
+          echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
+
+
+  release-cxx11-tarball-artifacts:
+    needs: [generate-release-tarball-matrix]
+    name: Release torch-tensorrt cxx11 tarball artifacts
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - repository: pytorch/tensorrt
+            package-name: torch_tensorrt
+            pre-script: packaging/pre_build_script.sh
+            env-var-script: packaging/env_vars.txt
+            post-script: packaging/post_build_script.sh
+            smoke-test-script: packaging/smoke_test_script.sh
+            is-release-tarball: true
+    uses: ./.github/workflows/build_wheels_linux.yml
+    with:
+      repository: ${{ matrix.repository }}
+      ref: ""
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      build-matrix: ${{ needs.generate-release-tarball-matrix.outputs.matrix }}
+      pre-script: ${{ matrix.pre-script }}
+      env-var-script: ${{ matrix.env-var-script }}
+      post-script: ${{ matrix.post-script }}
+      package-name: ${{ matrix.package-name }}
+      smoke-test-script: ${{ matrix.smoke-test-script }}
+      trigger-event: ${{ github.event_name }}
+      is-release-tarball: true
 
   release-wheel-artifacts:
-    needs: substitute-runner
+    needs: [generate-release-wheel-matrix]
     name: Release torch-tensorrt wheel artifacts
     strategy:
       fail-fast: false
@@ -69,17 +122,19 @@ jobs:
             env-script: packaging/vc_env_helper.bat
             smoke-test-script: packaging/smoke_test_windows.py
             package-name: torch_tensorrt
+            is-release-wheel: true
     uses: ./.github/workflows/release-wheel-windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
       test-infra-repository: pytorch/test-infra
       test-infra-ref: main
-      build-matrix: ${{ needs.substitute-runner.outputs.matrix }}
+      build-matrix: ${{ needs.generate-release-wheel-matrix.outputs.matrix }}
       pre-script: ${{ matrix.pre-script }}
       env-script: ${{ matrix.env-script }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       package-name: ${{ matrix.package-name }}
+      is-release-wheel: true
 
 
 concurrency:

From e3fc3ea8790e18292474e758b0a9e53dd5837090 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Thu, 28 Aug 2025 17:52:23 -0700
Subject: [PATCH 03/15] test CI

---
 .github/workflows/build-test-linux-aarch64-jetpack.yml | 2 +-
 .github/workflows/build-test-linux-aarch64.yml         | 2 +-
 .github/workflows/build-test-linux-x86_64.yml          | 2 +-
 .github/workflows/build-test-linux-x86_64_rtx.yml      | 2 +-
 .github/workflows/build-test-windows.yml               | 2 +-
 .github/workflows/build-test-windows_rtx.yml           | 2 +-
 6 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/build-test-linux-aarch64-jetpack.yml b/.github/workflows/build-test-linux-aarch64-jetpack.yml
index 9523872acf..3d4bcf50db 100644
--- a/.github/workflows/build-test-linux-aarch64-jetpack.yml
+++ b/.github/workflows/build-test-linux-aarch64-jetpack.yml
@@ -1,7 +1,7 @@
 name: Build and test Linux aarch64 wheels for Jetpack
 
 on:
-  pull_request:
+  #pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-linux-aarch64.yml b/.github/workflows/build-test-linux-aarch64.yml
index a2b2a78db9..3a353c3a1c 100644
--- a/.github/workflows/build-test-linux-aarch64.yml
+++ b/.github/workflows/build-test-linux-aarch64.yml
@@ -1,7 +1,7 @@
 name: Build and test Linux aarch64 wheels
 
 on:
-  pull_request:
+  #pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-linux-x86_64.yml b/.github/workflows/build-test-linux-x86_64.yml
index 3998eab0e7..5662810818 100644
--- a/.github/workflows/build-test-linux-x86_64.yml
+++ b/.github/workflows/build-test-linux-x86_64.yml
@@ -1,7 +1,7 @@
 name: Build and test Linux x86_64 wheels
 
 on:
-  pull_request:
+  #pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-linux-x86_64_rtx.yml b/.github/workflows/build-test-linux-x86_64_rtx.yml
index 1f9b605584..d278069b2d 100644
--- a/.github/workflows/build-test-linux-x86_64_rtx.yml
+++ b/.github/workflows/build-test-linux-x86_64_rtx.yml
@@ -1,7 +1,7 @@
 name: RTX - Build and test Linux x86_64 wheels
 
 on:
-  pull_request:
+  #pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-windows.yml b/.github/workflows/build-test-windows.yml
index ca39f8cb5f..647364e70e 100644
--- a/.github/workflows/build-test-windows.yml
+++ b/.github/workflows/build-test-windows.yml
@@ -1,7 +1,7 @@
 name: Build and test Windows wheels
 
 on:
-  pull_request:
+  #pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-windows_rtx.yml b/.github/workflows/build-test-windows_rtx.yml
index 137adb252c..3c401d1b21 100644
--- a/.github/workflows/build-test-windows_rtx.yml
+++ b/.github/workflows/build-test-windows_rtx.yml
@@ -1,7 +1,7 @@
 name: RTX - Build and test Windows wheels
 
 on:
-  pull_request:
+  #pull_request:
   push:
     branches:
       - main

From 98f6a7d9c83efcbbdd4d479d0109c058bf9e59fe Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 09:37:51 -0700
Subject: [PATCH 04/15] test

---
 .github/workflows/release-linux-aarch64.yml | 2 +-
 .github/workflows/release-linux-x86_64.yml  | 2 +-
 setup.py                                    | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index 85ab3672a9..46962d785f 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -127,5 +127,5 @@ jobs:
       is-release-wheel: true
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
+  group: ${{ github.workflow }}-aarch64-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
   cancel-in-progress: true
\ No newline at end of file
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index e0190c2a72..f38ad88d80 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -127,5 +127,5 @@ jobs:
       is-release-wheel: true
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
+  group: ${{ github.workflow }}-x86_64-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
   cancel-in-progress: true
\ No newline at end of file
diff --git a/setup.py b/setup.py
index b26a34f371..3ba8352f0b 100644
--- a/setup.py
+++ b/setup.py
@@ -631,7 +631,7 @@ def run(self):
             ),
             extra_compile_args=(
                 [
-                    f'/DPYBIND11_BUILD_ABI=\\"{torch._C._PYBIND11_BUILD_ABI}\\"',
+                    f'/DPYBIND11_BUILD_ABI=\\"{getattr(torch._C, "_PYBIND11_BUILD_ABI", "_cxxabi1018")}\\"',
                     "/GS-",
                     "/permissive-",
                     "/utf-8",

From f65e66a1bff64f0d8864f2daee60c5d583f6939a Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 09:42:04 -0700
Subject: [PATCH 05/15] test

---
 .github/workflows/release-linux-aarch64.yml | 2 ++
 .github/workflows/release-windows.yml       | 4 ++--
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index 46962d785f..d3f329da3a 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -75,6 +75,7 @@ jobs:
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
       is-release-tarball: true
+      architecture: "aarch64"
 
   generate-release-wheel-matrix:
     needs: [generate-matrix]
@@ -125,6 +126,7 @@ jobs:
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
       is-release-wheel: true
+      architecture: "aarch64"
 
 concurrency:
   group: ${{ github.workflow }}-aarch64-${{ github.event.pull_request.number || github.ref_name }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}-${{ inputs.job-name }}
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index bee7fc8721..a47b045ce5 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -95,7 +95,7 @@ jobs:
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
             is-release-tarball: true
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_wheels_windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
@@ -123,7 +123,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_windows.py
             package-name: torch_tensorrt
             is-release-wheel: true
-    uses: ./.github/workflows/release-wheel-windows.yml
+    uses: ./.github/workflows/build_wheels_windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""

From 1ccf39b315f53e921ece73ac6e3866f5d86278a8 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 10:00:29 -0700
Subject: [PATCH 06/15] add pkg_zip for windows zip release

---
 .github/workflows/release-linux-aarch64.yml |  1 -
 .github/workflows/release-linux-x86_64.yml  |  1 -
 .github/workflows/release-windows.yml       | 12 ++--
 BUILD.bazel                                 | 62 ++++++++++++++++++---
 4 files changed, 59 insertions(+), 17 deletions(-)

diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index d3f329da3a..d619f618a2 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -1,7 +1,6 @@
 name: Release aarch64 Linux wheels and tarball artifacts
 
 on:
-  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index f38ad88d80..320bd78a4a 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -1,7 +1,6 @@
 name: Release x86_64 Linux wheels and tarball artifacts
 
 on:
-  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index a47b045ce5..49dad0bcd8 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -1,7 +1,6 @@
 name: Release Windows wheels artifacts
 
 on:
-  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
@@ -90,10 +89,9 @@ jobs:
         include:
           - repository: pytorch/tensorrt
             package-name: torch_tensorrt
-            pre-script: packaging/pre_build_script.sh
-            env-var-script: packaging/env_vars.txt
-            post-script: packaging/post_build_script.sh
-            smoke-test-script: packaging/smoke_test_script.sh
+            pre-script: packaging/pre_build_script_windows.sh
+            env-script: packaging/vc_env_helper.bat
+            smoke-test-script: packaging/smoke_test_windows.py
             is-release-tarball: true
     uses: ./.github/workflows/build_wheels_windows.yml
     with:
@@ -103,11 +101,9 @@ jobs:
       test-infra-ref: main
       build-matrix: ${{ needs.generate-release-tarball-matrix.outputs.matrix }}
       pre-script: ${{ matrix.pre-script }}
-      env-var-script: ${{ matrix.env-var-script }}
-      post-script: ${{ matrix.post-script }}
+      env-script: ${{ matrix.env-var-script }}
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
-      trigger-event: ${{ github.event_name }}
       is-release-tarball: true
 
   release-wheel-artifacts:
diff --git a/BUILD.bazel b/BUILD.bazel
index 7a49d116b6..b70ec268bb 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -1,4 +1,4 @@
-load("@rules_pkg//:pkg.bzl", "pkg_tar")
+load("@rules_pkg//:pkg.bzl", "pkg_tar", "pkg_zip")
 
 config_setting(
     name = "windows",
@@ -96,8 +96,24 @@ pkg_tar(
     package_dir = "bin/",
 )
 
+# Windows-specific zip target
+pkg_zip(
+    name = "libtorchtrt_zip",
+    srcs = [
+        "//:LICENSE",
+        "//third_party/torch_tensorrt:BUILD",
+    ],
+    package_dir = "torch_tensorrt",
+    deps = [
+        ":include",
+        ":include_core",
+        ":lib",
+    ],
+)
+
+# Non-Windows tar target
 pkg_tar(
-    name = "libtorchtrt",
+    name = "libtorchtrt_tar",
     srcs = [
         "//:LICENSE",
         "//third_party/torch_tensorrt:BUILD",
@@ -105,18 +121,40 @@ pkg_tar(
     extension = "tar.gz",
     package_dir = "torch_tensorrt",
     deps = [
+        ":bin",
         ":include",
         ":include_core",
         ":lib",
-    ] + select({
-        ":rtx_win": [],
-        ":windows": [],
-        "//conditions:default": [":bin"],
+    ],
+)
+
+# Main target that selects between zip and tar based on platform
+alias(
+    name = "libtorchtrt",
+    actual = select({
+        ":rtx_win": ":libtorchtrt_zip",
+        ":windows": ":libtorchtrt_zip",
+        "//conditions:default": ":libtorchtrt_tar",
     }),
 )
 
+# Windows-specific zip target for runtime
+pkg_zip(
+    name = "libtorchtrt_runtime_zip",
+    srcs = [
+        "//:LICENSE",
+        "//third_party/torch_tensorrt:BUILD",
+    ],
+    package_dir = "torch_tensorrt_runtime",
+    deps = [
+        ":include_rt",
+        ":lib_rt",
+    ],
+)
+
+# Non-Windows tar target for runtime
 pkg_tar(
-    name = "libtorchtrt_runtime",
+    name = "libtorchtrt_runtime_tar",
     srcs = [
         "//:LICENSE",
         "//third_party/torch_tensorrt:BUILD",
@@ -128,3 +166,13 @@ pkg_tar(
         ":lib_rt",
     ],
 )
+
+# Main runtime target that selects between zip and tar based on platform
+alias(
+    name = "libtorchtrt_runtime",
+    actual = select({
+        ":rtx_win": ":libtorchtrt_runtime_zip",
+        ":windows": ":libtorchtrt_runtime_zip",
+        "//conditions:default": ":libtorchtrt_runtime_tar",
+    }),
+)

From a827667db27874a047e7435c82195481301425fa Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 11:07:59 -0700
Subject: [PATCH 07/15] add alias

---
 BUILD.bazel             | 94 +++++++++++++++++++++++++++++------------
 core/BUILD              | 11 ++++-
 core/plugins/BUILD      | 19 ++++++++-
 core/runtime/BUILD      | 15 ++++++-
 core/util/BUILD         | 16 ++++++-
 core/util/logging/BUILD |  9 +++-
 6 files changed, 133 insertions(+), 31 deletions(-)

diff --git a/BUILD.bazel b/BUILD.bazel
index b70ec268bb..5350ce5ac4 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -43,6 +43,20 @@ pkg_tar(
     ],
 )
 
+pkg_zip(
+    name = "include_core_zip",
+    srcs = [
+        "//core:include_zip",
+        "//core/plugins:impl_include_zip",
+        "//core/plugins:include_zip",
+        "//core/runtime:include_zip",
+        "//core/util:include_zip",
+        "//core/util/logging:include_zip",
+    ],
+    out = "include_core.zip",
+    package_dir = "include/torch_tensorrt",
+)
+
 pkg_tar(
     name = "include_rt",
     package_dir = "include/torch_tensorrt",
@@ -51,6 +65,15 @@ pkg_tar(
     ],
 )
 
+pkg_zip(
+    name = "include_rt_zip",
+    srcs = [
+        "//core/runtime:include",
+    ],
+    out = "include_rt.zip",
+    package_dir = "include/torch_tensorrt",
+)
+
 pkg_tar(
     name = "include",
     srcs = [
@@ -59,30 +82,49 @@ pkg_tar(
     package_dir = "include/torch_tensorrt/",
 )
 
+pkg_zip(
+    name = "include_zip",
+    srcs = [
+        "//cpp:api_headers",
+    ],
+    out = "include.zip",
+    package_dir = "include/torch_tensorrt/",
+)
+
+# Non-Windows tar target for lib
 pkg_tar(
     name = "lib",
-    srcs = select({
-        ":rtx_win": ["//cpp/lib:torchtrt.dll"],
-        ":windows": ["//cpp/lib:torchtrt.dll"],
-        "//conditions:default": [
-            "//cpp/lib:libtorchtrt.so",
-            "//cpp/lib:libtorchtrt_plugins.so",
-            "//cpp/lib:libtorchtrt_runtime.so",
-        ],
-    }),
+    srcs = [
+        "//cpp/lib:libtorchtrt.so",
+        "//cpp/lib:libtorchtrt_plugins.so",
+        "//cpp/lib:libtorchtrt_runtime.so",
+    ],
+    mode = "0755",
+    package_dir = "lib/",
+)
+
+# Windows-specific zip target for lib
+pkg_zip(
+    name = "lib_zip",
+    srcs = ["//cpp/lib:torchtrt.dll"],
+    out = "lib.zip",
     mode = "0755",
     package_dir = "lib/",
 )
 
+# Non-Windows tar target for lib_rt
 pkg_tar(
     name = "lib_rt",
-    srcs = select({
-        ":rtx_win": ["//cpp/lib:torchtrt_runtime.dll"],
-        ":windows": ["//cpp/lib:torchtrt_runtime.dll"],
-        "//conditions:default": [
-            "//cpp/lib:libtorchtrt_runtime.so",
-        ],
-    }),
+    srcs = ["//cpp/lib:libtorchtrt_runtime.so"],
+    mode = "0755",
+    package_dir = "lib/",
+)
+
+# Windows-specific zip target for lib_rt
+pkg_zip(
+    name = "lib_rt_zip",
+    srcs = ["//cpp/lib:torchtrt_runtime.dll"],
+    out = "lib_rt.zip",
     mode = "0755",
     package_dir = "lib/",
 )
@@ -100,15 +142,14 @@ pkg_tar(
 pkg_zip(
     name = "libtorchtrt_zip",
     srcs = [
+        ":include_core_zip",
+        ":include_zip",
+        ":lib_zip",
         "//:LICENSE",
         "//third_party/torch_tensorrt:BUILD",
     ],
+    out = "libtorchtrt.zip",
     package_dir = "torch_tensorrt",
-    deps = [
-        ":include",
-        ":include_core",
-        ":lib",
-    ],
 )
 
 # Non-Windows tar target
@@ -118,6 +159,7 @@ pkg_tar(
         "//:LICENSE",
         "//third_party/torch_tensorrt:BUILD",
     ],
+    out = "libtorchtrt.tar.gz",
     extension = "tar.gz",
     package_dir = "torch_tensorrt",
     deps = [
@@ -142,14 +184,13 @@ alias(
 pkg_zip(
     name = "libtorchtrt_runtime_zip",
     srcs = [
+        ":include_rt_zip",
+        ":lib_rt_zip",
         "//:LICENSE",
         "//third_party/torch_tensorrt:BUILD",
     ],
+    out = "libtorchtrt_runtime.zip",
     package_dir = "torch_tensorrt_runtime",
-    deps = [
-        ":include_rt",
-        ":lib_rt",
-    ],
 )
 
 # Non-Windows tar target for runtime
@@ -159,6 +200,7 @@ pkg_tar(
         "//:LICENSE",
         "//third_party/torch_tensorrt:BUILD",
     ],
+    out = "libtorchtrt_runtime.tar.gz",
     extension = "tar.gz",
     package_dir = "torch_tensorrt_runtime",
     deps = [
@@ -167,7 +209,7 @@ pkg_tar(
     ],
 )
 
-# Main runtime target that selects between zip and tar based on platform
+# Main target that selects between zip and tar based on platform
 alias(
     name = "libtorchtrt_runtime",
     actual = select({
diff --git a/core/BUILD b/core/BUILD
index c6744c66c1..2cd415f881 100644
--- a/core/BUILD
+++ b/core/BUILD
@@ -1,5 +1,5 @@
 load("@rules_cc//cc:defs.bzl", "cc_library")
-load("@rules_pkg//:pkg.bzl", "pkg_tar")
+load("@rules_pkg//:pkg.bzl", "pkg_tar", "pkg_zip")
 
 package(default_visibility = ["//visibility:public"])
 
@@ -103,3 +103,12 @@ pkg_tar(
     ],
     package_dir = "core/",
 )
+
+pkg_zip(
+    name = "include_zip",
+    srcs = [
+        "compiler.h",
+    ],
+    out = "include.zip",
+    package_dir = "core/",
+)
diff --git a/core/plugins/BUILD b/core/plugins/BUILD
index 00503552f2..ec32cc2ed5 100644
--- a/core/plugins/BUILD
+++ b/core/plugins/BUILD
@@ -1,5 +1,5 @@
 load("@rules_cc//cc:defs.bzl", "cc_library")
-load("@rules_pkg//:pkg.bzl", "pkg_tar")
+load("@rules_pkg//:pkg.bzl", "pkg_tar", "pkg_zip")
 
 package(default_visibility = ["//visibility:public"])
 
@@ -129,3 +129,20 @@ pkg_tar(
     ],
     package_dir = "core/plugins/impl",
 )
+
+pkg_zip(
+    name = "include_zip",
+    srcs = ["plugins.h"],
+    out = "include.zip",
+    package_dir = "core/plugins/",
+)
+
+pkg_zip(
+    name = "impl_include_zip",
+    srcs = [
+        "impl/interpolate_plugin.h",
+        "impl/normalize_plugin.h",
+    ],
+    out = "impl_include.zip",
+    package_dir = "core/plugins/impl",
+)
diff --git a/core/runtime/BUILD b/core/runtime/BUILD
index a573cfed78..18ab76f7c1 100644
--- a/core/runtime/BUILD
+++ b/core/runtime/BUILD
@@ -1,5 +1,5 @@
 load("@rules_cc//cc:defs.bzl", "cc_library")
-load("@rules_pkg//:pkg.bzl", "pkg_tar")
+load("@rules_pkg//:pkg.bzl", "pkg_tar", "pkg_zip")
 
 package(default_visibility = ["//visibility:public"])
 
@@ -111,3 +111,16 @@ pkg_tar(
     ],
     package_dir = "core/runtime/",
 )
+
+pkg_zip(
+    name = "include_zip",
+    srcs = [
+        "Platform.h",
+        "RTDevice.h",
+        "TRTEngine.h",
+        "TRTEngineProfiler.h",
+        "runtime.h",
+    ],
+    out = "include.zip",
+    package_dir = "core/runtime/",
+)
diff --git a/core/util/BUILD b/core/util/BUILD
index 0ed97a5eda..527bed4c04 100644
--- a/core/util/BUILD
+++ b/core/util/BUILD
@@ -1,5 +1,5 @@
 load("@rules_cc//cc:defs.bzl", "cc_library")
-load("@rules_pkg//:pkg.bzl", "pkg_tar")
+load("@rules_pkg//:pkg.bzl", "pkg_tar", "pkg_zip")
 
 package(default_visibility = ["//visibility:public"])
 
@@ -165,3 +165,17 @@ pkg_tar(
     ],
     package_dir = "core/util/",
 )
+
+pkg_zip(
+    name = "include_zip",
+    srcs = [
+        "//core/util:Exception.h",
+        "//core/util:build_info.h",
+        "//core/util:jit_util.h",
+        "//core/util:macros.h",
+        "//core/util:prelude.h",
+        "//core/util:trt_util.h",
+    ],
+    out = "include.zip",
+    package_dir = "core/util/",
+)
diff --git a/core/util/logging/BUILD b/core/util/logging/BUILD
index 1ac834b021..0d409df172 100644
--- a/core/util/logging/BUILD
+++ b/core/util/logging/BUILD
@@ -1,5 +1,5 @@
 load("@rules_cc//cc:defs.bzl", "cc_library")
-load("@rules_pkg//:pkg.bzl", "pkg_tar")
+load("@rules_pkg//:pkg.bzl", "pkg_tar", "pkg_zip")
 
 package(default_visibility = ["//visibility:public"])
 
@@ -85,3 +85,10 @@ pkg_tar(
     srcs = ["TorchTRTLogger.h"],
     package_dir = "core/util/logging",
 )
+
+pkg_zip(
+    name = "include_zip",
+    srcs = ["TorchTRTLogger.h"],
+    out = "include.zip",
+    package_dir = "core/util/logging",
+)

From c14654a137d5e655b2f3657a1da4a8490ec3a15f Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 11:26:48 -0700
Subject: [PATCH 08/15] test

---
 .github/workflows/release-linux-aarch64.yml | 1 +
 .github/workflows/release-linux-x86_64.yml  | 1 +
 .github/workflows/release-windows.yml       | 1 +
 3 files changed, 3 insertions(+)

diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index d619f618a2..d3f329da3a 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -1,6 +1,7 @@
 name: Release aarch64 Linux wheels and tarball artifacts
 
 on:
+  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index 320bd78a4a..f38ad88d80 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -1,6 +1,7 @@
 name: Release x86_64 Linux wheels and tarball artifacts
 
 on:
+  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index 49dad0bcd8..5273b9990e 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -1,6 +1,7 @@
 name: Release Windows wheels artifacts
 
 on:
+  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds

From 1a0f8cc124642ee1ae648a399a4ffef9cb77fafb Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 12:46:55 -0700
Subject: [PATCH 09/15] test

---
 .github/workflows/build_wheels_windows.yml  |  9 +++---
 .github/workflows/release-linux-aarch64.yml |  2 +-
 .github/workflows/release-windows.yml       | 32 +++++++++------------
 3 files changed, 20 insertions(+), 23 deletions(-)

diff --git a/.github/workflows/build_wheels_windows.yml b/.github/workflows/build_wheels_windows.yml
index 8eab386033..12735bed0e 100644
--- a/.github/workflows/build_wheels_windows.yml
+++ b/.github/workflows/build_wheels_windows.yml
@@ -295,8 +295,9 @@ jobs:
               mkdir -p release/tarball
               TRT_VERSION=$(cat dev_dep_versions.yml | grep __tensorrt_version__ | sed 's/__tensorrt_version__: //g' | sed 's/"//g')
               bazel build //:libtorchtrt --compilation_mode opt --config=windows
-              cp bazel-bin/libtorchtrt.tar.gz \
-              release/tarball/libtorchtrt-${BUILD_VERSION}-tensorrt${TRT_VERSION}-cuda${CU_VERSION:2}-libtorch${PYTORCH_VERSION}-x86_64-windows.tar.gz
+              ls -lart bazel-bin/
+              cp bazel-bin/libtorchtrt.zip\
+              release/tarball/libtorchtrt-${BUILD_VERSION}-tensorrt${TRT_VERSION}-cuda${CU_VERSION:2}-libtorch${PYTORCH_VERSION}-x86_64-windows.zip
               return 0
             fi
           fi
@@ -334,7 +335,7 @@ jobs:
           source "${BUILD_ENV_FILE}"
           ${CONDA_RUN} ${ENV_SCRIPT} ${POST_SCRIPT}
       - name: Smoke Test X64
-        if: inputs.architecture == 'x64'
+        if: ${{ inputs.architecture == 'x64' && inputs.is-release-tarball == false }}
         env:
           ENV_SCRIPT: ${{ inputs.env-script }}
           PACKAGE_NAME: ${{ inputs.package-name }}
@@ -421,5 +422,5 @@ jobs:
       architecture: ${{ inputs.architecture }}
 
 concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ inputs.repository }}-${{ github.event_name == 'workflow_dispatch' }}
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ inputs.repository }}-${{ inputs.is-release-wheel }}-${{ inputs.is-release-tarball }}-${{ github.event_name == 'workflow_dispatch' }}
   cancel-in-progress: true
\ No newline at end of file
diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index d3f329da3a..333a3208cd 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -20,7 +20,7 @@ jobs:
     if: ${{ contains(github.event.pull_request.labels.*.name, 'build-release-artifacts') || startsWith(github.event.ref, 'refs/tags/v') }}
     with:
       package-type: wheel
-      os: linux
+      os: linux-aarch64
       test-infra-repository: pytorch/test-infra
       test-infra-ref: main
       with-rocm: false
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index 5273b9990e..7116610983 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -26,20 +26,8 @@ jobs:
       with-rocm: false
       with-cpu: false
 
-  substitute-runner:
-    needs: generate-matrix
-    outputs:
-      matrix: ${{ steps.substitute.outputs.matrix }}
-    runs-on: ubuntu-latest
-    steps:
-      - name: Substitute runner
-        id: substitute
-        run: |
-          echo matrix="$(echo '${{ needs.generate-release-matrix.outputs.matrix }}' | sed -e 's/windows.g4dn.xlarge/windows.g5.4xlarge.nvidia.gpu/g')" >> ${GITHUB_OUTPUT}
-
-
   generate-release-tarball-matrix:
-    needs: [substitute-runner]
+    needs: [generate-matrix]
     outputs:
       matrix: ${{ steps.generate.outputs.matrix }}
     runs-on: ubuntu-latest
@@ -54,13 +42,16 @@ jobs:
         id: generate
         run: |
           set -eou pipefail
-          MATRIX_BLOB=${{ toJSON(needs.substitute-runner.outputs.matrix) }}
+          MATRIX_BLOB=${{ toJSON(needs.generate-matrix.outputs.matrix) }}
           MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --tarball_matrix "${MATRIX_BLOB}")"
           echo "${MATRIX_BLOB}"
-          echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
+          substitute_matrix="$(echo "${MATRIX_BLOB}" | sed -e 's/windows.g4dn.xlarge/windows.g5.4xlarge.nvidia.gpu/g')"
+
+          echo matrix="$(echo "${substitute_matrix}")" >> ${GITHUB_OUTPUT}
+
 
   generate-release-wheel-matrix:
-    needs: [substitute-runner]
+    needs: [generate-matrix]
     outputs:
       matrix: ${{ steps.generate.outputs.matrix }}
     runs-on: ubuntu-latest
@@ -75,10 +66,15 @@ jobs:
         id: generate
         run: |
           set -eou pipefail
-          MATRIX_BLOB=${{ toJSON(needs.substitute-runner.outputs.matrix) }}
+          MATRIX_BLOB=${{ toJSON(needs.generate-matrix.outputs.matrix) }}
           MATRIX_BLOB="$(python3 .github/scripts/generate-release-matrix.py --wheel_matrix "${MATRIX_BLOB}")"
+          echo "****MATRIX_BLOB****"
           echo "${MATRIX_BLOB}"
-          echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
+
+          substitute_matrix="$(echo "${MATRIX_BLOB}" | sed -e 's/windows.g4dn.xlarge/windows.g5.4xlarge.nvidia.gpu/g')"
+          echo "****substitute_matrix****"
+          echo "${substitute_matrix}"
+          echo "matrix=${substitute_matrix}" >> "${GITHUB_OUTPUT}"
 
 
   release-cxx11-tarball-artifacts:

From e6f26570f6a01534c1f596d1a4e774b2939477eb Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 13:19:46 -0700
Subject: [PATCH 10/15] sbsa use sbsa container image for release build

---
 .github/scripts/generate-release-matrix.py | 11 +++++++----
 third_party/libtorch/BUILD                 |  1 +
 2 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/.github/scripts/generate-release-matrix.py b/.github/scripts/generate-release-matrix.py
index 7b9e0263dd..9f740c2f4f 100644
--- a/.github/scripts/generate-release-matrix.py
+++ b/.github/scripts/generate-release-matrix.py
@@ -12,6 +12,7 @@
     "wheel": ["3.10", "3.11", "3.12", "3.13"],
     "tarball": ["3.11"],
 }
+sbsa_container_image: str = "quay.io/pypa/manylinux_2_34_aarch64"
 
 CXX11_TARBALL_CONTAINER_IMAGE = {
     "cu129": "pytorch/libtorch-cxx11-builder:cuda12.9-main",
@@ -56,10 +57,12 @@ def main(args: list[str]) -> None:
             item["desired_cuda"] in cuda_versions
             and item["python_version"] in python_versions
         ):
-            if options.tarball_matrix != "":
-                item["container_image"] = CXX11_TARBALL_CONTAINER_IMAGE[
-                    item["desired_cuda"]
-                ]
+            if item["gpu_arch_type"] == "cuda-aarch64":
+                item["container_image"] = sbsa_container_image
+            # if options.tarball_matrix != "":
+            #     item["container_image"] = CXX11_TARBALL_CONTAINER_IMAGE[
+            #         item["desired_cuda"]
+            #     ]
             filtered_includes.append(item)
     filtered_matrix_dict = {}
     filtered_matrix_dict["include"] = filtered_includes
diff --git a/third_party/libtorch/BUILD b/third_party/libtorch/BUILD
index 46de7dd8d0..37309f7209 100644
--- a/third_party/libtorch/BUILD
+++ b/third_party/libtorch/BUILD
@@ -35,6 +35,7 @@ cc_library(
         [
             "include/torch/**/*.h",
         ],
+        allow_empty = True,
         exclude = [
             "include/torch/csrc/api/include/**/*.h",
         ],

From 260ca74e6d7cd5f60ba56469fa59dab3bb180058 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 13:51:01 -0700
Subject: [PATCH 11/15] rename zip tarball name

---
 .github/workflows/build_wheels_linux.yml    |  4 ++--
 .github/workflows/build_wheels_windows.yml  |  4 ++--
 .github/workflows/release-linux-aarch64.yml |  4 ++--
 .github/workflows/release-linux-x86_64.yml  |  4 ++--
 .github/workflows/release-windows.yml       | 12 ++++++------
 5 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/.github/workflows/build_wheels_linux.yml b/.github/workflows/build_wheels_linux.yml
index bdcff46256..32dd21de82 100644
--- a/.github/workflows/build_wheels_linux.yml
+++ b/.github/workflows/build_wheels_linux.yml
@@ -385,12 +385,12 @@ jobs:
         with:
           name: ${{ env.ARTIFACT_NAME }}
           path: ${{ inputs.repository }}/dist/
-      - name: Upload tarball to GitHub
+      - name: Upload cxx11 tarball to GitHub
         if: ${{ inputs.is-release-tarball == true }}
         continue-on-error: true
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.ARTIFACT_NAME }}
+          name: cxx11-tarball-${{ env.PYTHON_VERSION }}-${{ env.CU_VERSION }}
           path: ${{ inputs.repository }}/release/tarball/
 
   upload:
diff --git a/.github/workflows/build_wheels_windows.yml b/.github/workflows/build_wheels_windows.yml
index 12735bed0e..6fbd343fdb 100644
--- a/.github/workflows/build_wheels_windows.yml
+++ b/.github/workflows/build_wheels_windows.yml
@@ -395,12 +395,12 @@ jobs:
         with:
           name: ${{ env.ARTIFACT_NAME }}
           path: ${{ inputs.repository }}/dist/
-      - name: Upload tarball to GitHub
+      - name: Upload zip artifact to GitHub
         if: ${{ inputs.is-release-tarball == true }}
         continue-on-error: true
         uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.ARTIFACT_NAME }}
+          name: cxx11-zip-${{ env.PYTHON_VERSION }}-${{ env.CU_VERSION }}
           path: ${{ inputs.repository }}/release/tarball/
       - uses: ./test-infra/.github/actions/teardown-windows
         if: inputs.architecture == 'x64'
diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index 333a3208cd..8a80050fba 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -49,7 +49,7 @@ jobs:
 
   release-cxx11-tarball-artifacts:
     needs: [generate-release-tarball-matrix]
-    name: Release torch-tensorrt cxx11 tarball artifacts
+    name: Release aarch64 torch-tensorrt cxx11 tarball artifacts
     strategy:
       fail-fast: false
       matrix:
@@ -99,7 +99,7 @@ jobs:
           echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
 
   release-wheel-artifacts:
-    name: Release torch-tensorrt wheel artifacts
+    name: Release aarch64 torch-tensorrt wheel artifacts
     needs: [generate-release-wheel-matrix]
     strategy:
       fail-fast: false
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index f38ad88d80..e98ecdb848 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -49,7 +49,7 @@ jobs:
 
   release-cxx11-tarball-artifacts:
     needs: [generate-release-tarball-matrix]
-    name: Release torch-tensorrt cxx11 tarball artifacts
+    name: Release x86_64 torch-tensorrt cxx11 tarball artifacts
     strategy:
       fail-fast: false
       matrix:
@@ -98,7 +98,7 @@ jobs:
           echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
 
   release-wheel-artifacts:
-    name: Release torch-tensorrt wheel artifacts
+    name: Release x86_64 torch-tensorrt wheel artifacts
     needs: [generate-release-wheel-matrix]
     strategy:
       fail-fast: false
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index 7116610983..a12e437712 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -26,7 +26,7 @@ jobs:
       with-rocm: false
       with-cpu: false
 
-  generate-release-tarball-matrix:
+  generate-release-zip-matrix:
     needs: [generate-matrix]
     outputs:
       matrix: ${{ steps.generate.outputs.matrix }}
@@ -77,9 +77,9 @@ jobs:
           echo "matrix=${substitute_matrix}" >> "${GITHUB_OUTPUT}"
 
 
-  release-cxx11-tarball-artifacts:
-    needs: [generate-release-tarball-matrix]
-    name: Release torch-tensorrt cxx11 tarball artifacts
+  release-cxx11-zip-artifacts:
+    needs: [generate-release-zip-matrix]
+    name: Release Windows torch-tensorrt cxx11 zip artifacts
     strategy:
       fail-fast: false
       matrix:
@@ -96,7 +96,7 @@ jobs:
       ref: ""
       test-infra-repository: pytorch/test-infra
       test-infra-ref: main
-      build-matrix: ${{ needs.generate-release-tarball-matrix.outputs.matrix }}
+      build-matrix: ${{ needs.generate-release-zip-matrix.outputs.matrix }}
       pre-script: ${{ matrix.pre-script }}
       env-script: ${{ matrix.env-var-script }}
       package-name: ${{ matrix.package-name }}
@@ -105,7 +105,7 @@ jobs:
 
   release-wheel-artifacts:
     needs: [generate-release-wheel-matrix]
-    name: Release torch-tensorrt wheel artifacts
+    name: Release Windows torch-tensorrt wheel artifacts
     strategy:
       fail-fast: false
       matrix:

From 4c2f4efbbe183d968415d2a6d49086e06171d3e2 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Fri, 29 Aug 2025 16:28:09 -0700
Subject: [PATCH 12/15] test

---
 BUILD.bazel | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/BUILD.bazel b/BUILD.bazel
index 5350ce5ac4..3d42625deb 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -152,7 +152,7 @@ pkg_zip(
     package_dir = "torch_tensorrt",
 )
 
-# Non-Windows tar target
+# tar target
 pkg_tar(
     name = "libtorchtrt_tar",
     srcs = [
@@ -163,11 +163,14 @@ pkg_tar(
     extension = "tar.gz",
     package_dir = "torch_tensorrt",
     deps = [
-        ":bin",
         ":include",
         ":include_core",
         ":lib",
-    ],
+    ] + select({
+        ":rtx_win": [],
+        ":windows": [],
+        "//conditions:default": [":bin"],
+    }),
 )
 
 # Main target that selects between zip and tar based on platform

From 6238fb569026e67874c4572072722b90607e3cfd Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Tue, 2 Sep 2025 08:32:32 -0700
Subject: [PATCH 13/15] test

---
 .github/workflows/build-test-linux-aarch64-jetpack.yml | 2 +-
 .github/workflows/build-test-linux-aarch64.yml         | 2 +-
 .github/workflows/build-test-linux-x86_64.yml          | 2 +-
 .github/workflows/build-test-linux-x86_64_rtx.yml      | 2 +-
 .github/workflows/build-test-windows.yml               | 2 +-
 .github/workflows/build-test-windows_rtx.yml           | 2 +-
 .github/workflows/release-linux-aarch64.yml            | 1 -
 .github/workflows/release-linux-x86_64.yml             | 1 -
 .github/workflows/release-windows.yml                  | 1 -
 9 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/.github/workflows/build-test-linux-aarch64-jetpack.yml b/.github/workflows/build-test-linux-aarch64-jetpack.yml
index 3d4bcf50db..9523872acf 100644
--- a/.github/workflows/build-test-linux-aarch64-jetpack.yml
+++ b/.github/workflows/build-test-linux-aarch64-jetpack.yml
@@ -1,7 +1,7 @@
 name: Build and test Linux aarch64 wheels for Jetpack
 
 on:
-  #pull_request:
+  pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-linux-aarch64.yml b/.github/workflows/build-test-linux-aarch64.yml
index 3a353c3a1c..a2b2a78db9 100644
--- a/.github/workflows/build-test-linux-aarch64.yml
+++ b/.github/workflows/build-test-linux-aarch64.yml
@@ -1,7 +1,7 @@
 name: Build and test Linux aarch64 wheels
 
 on:
-  #pull_request:
+  pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-linux-x86_64.yml b/.github/workflows/build-test-linux-x86_64.yml
index 5662810818..3998eab0e7 100644
--- a/.github/workflows/build-test-linux-x86_64.yml
+++ b/.github/workflows/build-test-linux-x86_64.yml
@@ -1,7 +1,7 @@
 name: Build and test Linux x86_64 wheels
 
 on:
-  #pull_request:
+  pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-linux-x86_64_rtx.yml b/.github/workflows/build-test-linux-x86_64_rtx.yml
index d278069b2d..1f9b605584 100644
--- a/.github/workflows/build-test-linux-x86_64_rtx.yml
+++ b/.github/workflows/build-test-linux-x86_64_rtx.yml
@@ -1,7 +1,7 @@
 name: RTX - Build and test Linux x86_64 wheels
 
 on:
-  #pull_request:
+  pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-windows.yml b/.github/workflows/build-test-windows.yml
index 647364e70e..ca39f8cb5f 100644
--- a/.github/workflows/build-test-windows.yml
+++ b/.github/workflows/build-test-windows.yml
@@ -1,7 +1,7 @@
 name: Build and test Windows wheels
 
 on:
-  #pull_request:
+  pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/build-test-windows_rtx.yml b/.github/workflows/build-test-windows_rtx.yml
index 3c401d1b21..137adb252c 100644
--- a/.github/workflows/build-test-windows_rtx.yml
+++ b/.github/workflows/build-test-windows_rtx.yml
@@ -1,7 +1,7 @@
 name: RTX - Build and test Windows wheels
 
 on:
-  #pull_request:
+  pull_request:
   push:
     branches:
       - main
diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index 8a80050fba..8b3190124b 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -1,7 +1,6 @@
 name: Release aarch64 Linux wheels and tarball artifacts
 
 on:
-  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index e98ecdb848..cbffc520ad 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -1,7 +1,6 @@
 name: Release x86_64 Linux wheels and tarball artifacts
 
 on:
-  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index a12e437712..590fd00fe9 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -1,7 +1,6 @@
 name: Release Windows wheels artifacts
 
 on:
-  pull_request:
   push:
     tags:
       # NOTE: Binary build pipelines should only get triggered on release candidate builds

From b7ca962b93ceb58c1fb5f3d1203179c1076bc157 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Wed, 3 Sep 2025 08:57:08 -0700
Subject: [PATCH 14/15] add filegroup to remove duplicated file lists in
 pkg_tar and pkg_zip

---
 core/plugins/BUILD | 31 ++++++++++++++++++++-----------
 core/runtime/BUILD | 18 +++++++++---------
 core/util/BUILD    | 31 +++++++++++++++----------------
 3 files changed, 44 insertions(+), 36 deletions(-)

diff --git a/core/plugins/BUILD b/core/plugins/BUILD
index ec32cc2ed5..37a9a47003 100644
--- a/core/plugins/BUILD
+++ b/core/plugins/BUILD
@@ -115,19 +115,16 @@ cc_library(
     alwayslink = True,
 )
 
-pkg_tar(
-    name = "include",
+filegroup(
+    name = "include_files",
     srcs = ["plugins.h"],
-    package_dir = "core/plugins/",
+    visibility = ["//visibility:public"],
 )
 
 pkg_tar(
-    name = "impl_include",
-    srcs = [
-        "impl/interpolate_plugin.h",
-        "impl/normalize_plugin.h",
-    ],
-    package_dir = "core/plugins/impl",
+    name = "include",
+    srcs = [":include_files"],
+    package_dir = "core/plugins/",
 )
 
 pkg_zip(
@@ -137,12 +134,24 @@ pkg_zip(
     package_dir = "core/plugins/",
 )
 
-pkg_zip(
-    name = "impl_include_zip",
+filegroup(
+    name = "impl_include_files",
     srcs = [
         "impl/interpolate_plugin.h",
         "impl/normalize_plugin.h",
     ],
+    visibility = ["//visibility:public"],
+)
+
+pkg_tar(
+    name = "impl_include",
+    srcs = [":impl_include_files"],
+    package_dir = "core/plugins/impl",
+)
+
+pkg_zip(
+    name = "impl_include_zip",
+    srcs = [":impl_include_files"],
     out = "impl_include.zip",
     package_dir = "core/plugins/impl",
 )
diff --git a/core/runtime/BUILD b/core/runtime/BUILD
index 18ab76f7c1..95576aa18b 100644
--- a/core/runtime/BUILD
+++ b/core/runtime/BUILD
@@ -100,8 +100,8 @@ cc_library(
     alwayslink = True,
 )
 
-pkg_tar(
-    name = "include",
+filegroup(
+    name = "include_files",
     srcs = [
         "Platform.h",
         "RTDevice.h",
@@ -109,18 +109,18 @@ pkg_tar(
         "TRTEngineProfiler.h",
         "runtime.h",
     ],
+    visibility = ["//visibility:public"],
+)
+
+pkg_tar(
+    name = "include",
+    srcs = [":include_files"],
     package_dir = "core/runtime/",
 )
 
 pkg_zip(
     name = "include_zip",
-    srcs = [
-        "Platform.h",
-        "RTDevice.h",
-        "TRTEngine.h",
-        "TRTEngineProfiler.h",
-        "runtime.h",
-    ],
+    srcs = [":include_files"],
     out = "include.zip",
     package_dir = "core/runtime/",
 )
diff --git a/core/util/BUILD b/core/util/BUILD
index 527bed4c04..8245e1c14a 100644
--- a/core/util/BUILD
+++ b/core/util/BUILD
@@ -153,29 +153,28 @@ cc_library(
     alwayslink = True,
 )
 
-pkg_tar(
-    name = "include",
+filegroup(
+    name = "include_files",
     srcs = [
-        "//core/util:Exception.h",
-        "//core/util:build_info.h",
-        "//core/util:jit_util.h",
-        "//core/util:macros.h",
-        "//core/util:prelude.h",
-        "//core/util:trt_util.h",
+        "Exception.h",
+        "build_info.h",
+        "jit_util.h",
+        "macros.h",
+        "prelude.h",
+        "trt_util.h",
     ],
+    visibility = ["//visibility:public"],
+)
+
+pkg_tar(
+    name = "include",
+    srcs = ["//core/util:include_files"],
     package_dir = "core/util/",
 )
 
 pkg_zip(
     name = "include_zip",
-    srcs = [
-        "//core/util:Exception.h",
-        "//core/util:build_info.h",
-        "//core/util:jit_util.h",
-        "//core/util:macros.h",
-        "//core/util:prelude.h",
-        "//core/util:trt_util.h",
-    ],
+    srcs = ["//core/util:include_files"],
     out = "include.zip",
     package_dir = "core/util/",
 )

From 51f0682cb1e308c4bdbe10b2b0fa10d719c7e7f1 Mon Sep 17 00:00:00 2001
From: lanluo-nvidia <lanl@nvidia.com>
Date: Wed, 3 Sep 2025 17:44:25 -0700
Subject: [PATCH 15/15] rename build_wheels_linux|windows.yml to
 build_linux|windows.yml

---
 .github/workflows/build-test-linux-aarch64-jetpack.yml        | 2 +-
 .github/workflows/build-test-linux-aarch64.yml                | 2 +-
 .github/workflows/build-test-linux-x86_64.yml                 | 2 +-
 .github/workflows/build-test-linux-x86_64_rtx.yml             | 2 +-
 .github/workflows/build-test-windows.yml                      | 2 +-
 .github/workflows/build-test-windows_rtx.yml                  | 2 +-
 .github/workflows/{build_wheels_linux.yml => build_linux.yml} | 0
 .../workflows/{build_wheels_windows.yml => build_windows.yml} | 0
 .github/workflows/release-linux-aarch64.yml                   | 4 ++--
 .github/workflows/release-linux-x86_64.yml                    | 4 ++--
 .github/workflows/release-windows.yml                         | 4 ++--
 11 files changed, 12 insertions(+), 12 deletions(-)
 rename .github/workflows/{build_wheels_linux.yml => build_linux.yml} (100%)
 rename .github/workflows/{build_wheels_windows.yml => build_windows.yml} (100%)

diff --git a/.github/workflows/build-test-linux-aarch64-jetpack.yml b/.github/workflows/build-test-linux-aarch64-jetpack.yml
index 9523872acf..ff827ad3f6 100644
--- a/.github/workflows/build-test-linux-aarch64-jetpack.yml
+++ b/.github/workflows/build-test-linux-aarch64-jetpack.yml
@@ -66,7 +66,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_script.sh
             package-name: torch_tensorrt
     name: Build torch-tensorrt whl package for jetpack
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/build-test-linux-aarch64.yml b/.github/workflows/build-test-linux-aarch64.yml
index a2b2a78db9..2604d18f92 100644
--- a/.github/workflows/build-test-linux-aarch64.yml
+++ b/.github/workflows/build-test-linux-aarch64.yml
@@ -63,7 +63,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_script.sh
             package-name: torch_tensorrt
     name: Build torch-tensorrt whl package for SBSA
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/build-test-linux-x86_64.yml b/.github/workflows/build-test-linux-x86_64.yml
index 3998eab0e7..b1630c03be 100644
--- a/.github/workflows/build-test-linux-x86_64.yml
+++ b/.github/workflows/build-test-linux-x86_64.yml
@@ -62,7 +62,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_script.sh
             package-name: torch_tensorrt
     name: Build torch-tensorrt whl package for Linux x86_64
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/build-test-linux-x86_64_rtx.yml b/.github/workflows/build-test-linux-x86_64_rtx.yml
index 1f9b605584..67eecef0e9 100644
--- a/.github/workflows/build-test-linux-x86_64_rtx.yml
+++ b/.github/workflows/build-test-linux-x86_64_rtx.yml
@@ -61,7 +61,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_script.sh
             package-name: torch_tensorrt
     name: RTX - Build torch-tensorrt-rtx whl package for Linux x86_64
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/build-test-windows.yml b/.github/workflows/build-test-windows.yml
index ca39f8cb5f..39c3d20571 100644
--- a/.github/workflows/build-test-windows.yml
+++ b/.github/workflows/build-test-windows.yml
@@ -71,7 +71,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_windows.py
             package-name: torch_tensorrt
     name: Build torch-tensorrt whl package for Windows
-    uses: ./.github/workflows/build_wheels_windows.yml
+    uses: ./.github/workflows/build_windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/build-test-windows_rtx.yml b/.github/workflows/build-test-windows_rtx.yml
index 137adb252c..88d347501d 100644
--- a/.github/workflows/build-test-windows_rtx.yml
+++ b/.github/workflows/build-test-windows_rtx.yml
@@ -71,7 +71,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_windows.py
             package-name: torch_tensorrt
     name: RTX - Build torch-tensorrt-rtx whl package for Windows
-    uses: ./.github/workflows/build_wheels_windows.yml
+    uses: ./.github/workflows/build_windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/build_wheels_linux.yml b/.github/workflows/build_linux.yml
similarity index 100%
rename from .github/workflows/build_wheels_linux.yml
rename to .github/workflows/build_linux.yml
diff --git a/.github/workflows/build_wheels_windows.yml b/.github/workflows/build_windows.yml
similarity index 100%
rename from .github/workflows/build_wheels_windows.yml
rename to .github/workflows/build_windows.yml
diff --git a/.github/workflows/release-linux-aarch64.yml b/.github/workflows/release-linux-aarch64.yml
index 8b3190124b..065962ab3a 100644
--- a/.github/workflows/release-linux-aarch64.yml
+++ b/.github/workflows/release-linux-aarch64.yml
@@ -60,7 +60,7 @@ jobs:
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
             is-release-tarball: "true"
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
@@ -111,7 +111,7 @@ jobs:
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
             is-release-wheel:  true
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/release-linux-x86_64.yml b/.github/workflows/release-linux-x86_64.yml
index cbffc520ad..fae96128c9 100644
--- a/.github/workflows/release-linux-x86_64.yml
+++ b/.github/workflows/release-linux-x86_64.yml
@@ -60,7 +60,7 @@ jobs:
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
             is-release-tarball: "true"
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
@@ -110,7 +110,7 @@ jobs:
             post-script: packaging/post_build_script.sh
             smoke-test-script: packaging/smoke_test_script.sh
             is-release-wheel:  true
-    uses: ./.github/workflows/build_wheels_linux.yml
+    uses: ./.github/workflows/build_linux.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
diff --git a/.github/workflows/release-windows.yml b/.github/workflows/release-windows.yml
index 590fd00fe9..fc70b83f96 100644
--- a/.github/workflows/release-windows.yml
+++ b/.github/workflows/release-windows.yml
@@ -89,7 +89,7 @@ jobs:
             env-script: packaging/vc_env_helper.bat
             smoke-test-script: packaging/smoke_test_windows.py
             is-release-tarball: true
-    uses: ./.github/workflows/build_wheels_windows.yml
+    uses: ./.github/workflows/build_windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""
@@ -115,7 +115,7 @@ jobs:
             smoke-test-script: packaging/smoke_test_windows.py
             package-name: torch_tensorrt
             is-release-wheel: true
-    uses: ./.github/workflows/build_wheels_windows.yml
+    uses: ./.github/workflows/build_windows.yml
     with:
       repository: ${{ matrix.repository }}
       ref: ""