infra: Testing out GHA based infra

Signed-off-by: Naren Dasan <naren@narendasan.com>
Signed-off-by: Naren Dasan <narens@nvidia.com>

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

@@ -0,0 +1,53 @@
+name: Build Linux Wheels
+
+on:
+  pull_request:
+  push:
+    branches:
+      - nightly
+      - main
+      - release/*
+    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:
+
+jobs:
+  generate-matrix:
+    uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
+    with:
+      package-type: wheel
+      os: linux
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      with-rocm: false
+  build:
+    needs: generate-matrix
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - repository: pytorch/tensorrt
+            pre-script: packaging/pre_build_script.sh
+            env-var-script: packaging/env_vars.txt
+            post-script: ""
+            smoke-test-script: ""
+            package-name: torch_tensorrt
+    name: ${{ matrix.repository }}
+    uses: pytorch/test-infra/.github/workflows/build_wheels_linux.yml@main
+    with:
+      repository: ${{ matrix.repository }}
+      ref: ""
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      build-matrix: ${{ needs.generate-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 }}
+    secrets:
+      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
+      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/docgen.yml

@@ -10,32 +10,33 @@ on:
 
 jobs:
   build-docs:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-latest
     container:
       image: docker.io/pytorch/manylinux-builder:cuda12.1
     steps:
       - uses: actions/checkout@v3
         with:
           ref: ${{github.head_ref}}
+      - name: Select python
+        run: |
+          export PATH=/opt/python/cp311-cp311/bin/:$PATH
       - name: Install base deps
         run: |
-          ./packaging/pre_build_script.sh
+          python3 -m pip install pyyaml
+          VERSION_SUFFIX=cu121 ./packaging/pre_build_script.sh
       - name: Get HEAD SHA
         id: vars
         run: echo "sha=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
-      - name: Get Bazel version
-        id: bazel_info
-        run: echo "version=$(cat .bazelversion)" >> $GITHUB_OUTPUT
       - name: Build Python Package
         run: |
-          cp toolchains/ci_workspaces/WORKSPACE.x86_64.cu121.release.rhel WORKSPACE
-          python -m pip install pip<=23
-          python -m pip install --pre -e . --extra-index-url https://download.pytorch.org/whl/nightly/cu121
+          python3 -m pip install pip --upgrade
+          CUDA_HOME=/usr/local/cuda-12.1 CI_BUILD=1 python3 -m pip install --pre -e . --extra-index-url https://download.pytorch.org/whl/nightly/cu121
       - name: Generate New Docs
         run: |
           cd docsrc
-          python -m pip install -r requirements.txt
-          python -c "import torch_tensorrt; print(torch_tensorrt.__version__)"
+          yum install -y doxygen pandoc
+          python3 -m pip install -r requirements.txt
+          python3 -c "import torch_tensorrt; print(torch_tensorrt.__version__)"
           make html
           cd ..
       - uses: stefanzweifel/git-auto-commit-action@v4

.github/workflows/nightlies.yml

@@ -0,0 +1,27 @@
+name: Cut Nightly
+
+on:
+  schedule:
+    - cron: '30 9 * * *'
+  workflow_dispatch:
+
+jobs:
+  cut_nightly:
+    runs-on: ubuntu-latest
+    timeout-minutes: 60
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          ref: main
+      - name: Create nightly commit
+        shell: bash
+        run: |
+          git config --global user.email "torch-tensorrt.github.bot@nvidia.com"
+          git config --global user.name "Torch-TensorRT Github Bot"
+          git fetch origin nightly
+          HEAD_COMMIT_HASH=$(git rev-parse HEAD)
+          NIGHTLY_DATE=$(date +"%Y-%m-%d")
+          # shellcheck disable=SC1083
+          NIGHTLY_RELEASE_COMMIT=$(git commit-tree -p FETCH_HEAD HEAD^{tree} -m "${NIGHTLY_DATE} nightly release (${HEAD_COMMIT_HASH})")
+          # shellcheck disable=SC1083
+          git push -f origin "${NIGHTLY_RELEASE_COMMIT}:nightly"

.github/workflows/tests.yml

@@ -0,0 +1,49 @@
+name: Tests for Torch-TensorRT
+
+on:
+  pull_request:
+  push:
+    branches:
+      - nightly
+      - main
+      - release/*
+    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:
+
+jobs:
+  generate-matrix:
+    uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
+    with:
+      package-type: wheel
+      os: linux
+      test-infra-repository: pytorch/test-infra
+      test-infra-ref: main
+      with-rocm: false
+
+  test-torchscript:
+    needs: generate-matrix
+    strategy:
+        fail-fast: false
+        matrix: ${{ fromJSON(needs.generate-matrix.outputs.matrix) }}
+    env:
+        PYTHON_VERSION: ${{ matrix.python_version }}
+        PACKAGE_TYPE: wheel
+        REPOSITORY: pytorch/tensorrt
+        REF: ""
+        CU_VERSION: ${{ matrix.desired_cuda }}
+    name: ${{ 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 60 minutes on these runners we need
+    # to have a conversation
+    timeout-minutes: 60
+    steps:
+        - name: Clean workspace
+          run: |
+            printenv
+

packaging/env_vars.txt

@@ -0,0 +1,2 @@
+export CI_BUILD="1"
+export RELEASE="1"

packaging/pre_build_script.sh

@@ -1,13 +1,16 @@
 #!/bin/bash
 
 # Install dependencies
+python3 -m pip install pyyaml
 TRT_VERSION=$(python3 -c "import versions; versions.tensorrt_version()")
 yum-config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo
 yum check-update
-yum install -y ninja-build tensorrt-${TRT_VERSION}.*
+yum install -y ninja-build gettext tensorrt-${TRT_VERSION}.*
 wget https://github.com/bazelbuild/bazelisk/releases/download/v1.17.0/bazelisk-linux-amd64 \
     && mv bazelisk-linux-amd64 /usr/bin/bazel \
     && chmod +x /usr/bin/bazel
 
-cp toolchains/ci_workspaces/WORKSPACE.x86_64.${VERSION_SUFFIX#*+}.release.rhel WORKSPACE
+export TORCH_BUILD_NUMBER=$(python -c "import torch, urllib.parse as ul; print(ul.quote_plus(torch.__version__))")
+
+cat toolchains/ci_workspaces/WORKSPACE.x86_64.release.rhel.tmpl | envsubst > WORKSPACE
 export CI_BUILD=1

py/ci/build_whl.sh

@@ -6,11 +6,23 @@ export CXX=g++
 export CUDA_HOME=/usr/local/cuda-12.1
 export PROJECT_DIR=/workspace/project
 
-cp -r $CUDA_HOME /usr/local/cuda
+rm -rf /usr/local/cuda
+
+if [[ $CUDA_HOME == "/usr/local/cuda-12.1" ]]; then
+    cp -r /usr/local/cuda-11.8 /usr/local/cuda
+    cp -r /usr/local/cuda-12.0/ /usr/local/cuda/
+    rsync -a /usr/local/cuda-12.1/ /usr/local/cuda/
+    export CUDA_HOME=/usr/local/cuda
+else
+    ln -s $CUDA_HOME /usr/local/cuda
+fi
 
 build_wheel() {
-    $1/bin/python -m pip install --upgrade pip
-    $1/bin/python -m pip wheel . --config-setting="--build-option=--release" --config-setting="--build-option=--ci" -w dist
+    $1/bin/python -m pip install --upgrade pip setuptools
+    $1/bin/python -m pip install -r py/requirements.txt
+    #$1/bin/python -m pip wheel . -w dist
+    export BUILD_VERSION=$(cd ${PROJECT_DIR} && $1/bin/python3 -c "import versions; versions.torch_tensorrt_version_release()")
+    CI_BUILD=1 RELEASE=1 $1/bin/python setup.py bdist_wheel
 }
 
 patch_wheel() {
@@ -86,7 +98,7 @@ libtorchtrt() {
     ${PY_DIR}/bin/python -m pip install setuptools wheel auditwheel
     bazel build //:libtorchtrt --platforms //toolchains:ci_rhel_x86_64_linux -c opt --noshow_progress
     CUDA_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cuda_version()")
-    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version()")
+    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version_release()")
     TRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.tensorrt_version()")
     CUDNN_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cudnn_version()")
     TORCH_VERSION=$(${PY_DIR}/bin/python -c "from torch import __version__;print(__version__.split('+')[0])")
@@ -106,7 +118,7 @@ libtorchtrt_pre_cxx11_abi() {
     ${PY_DIR}/bin/python -m pip install setuptools wheel auditwheel
     bazel build //:libtorchtrt --config pre_cxx11_abi --platforms //toolchains:ci_rhel_x86_64_linux -c opt --noshow_progress
     CUDA_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cuda_version()")
-    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version()")
+    TORCHTRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.torch_tensorrt_version_release()")
     TRT_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.tensorrt_version()")
     CUDNN_VERSION=$(cd ${PROJECT_DIR} && ${PY_DIR}/bin/python3 -c "import versions; versions.cudnn_version()")
     TORCH_VERSION=$(${PY_DIR}/bin/python -c "from torch import __version__;print(__version__.split('+')[0])")

pyproject.toml

@@ -9,8 +9,8 @@ requires = [
     "typing-extensions>=4.7.0",
     "future>=0.18.3",
     "tensorrt>=8.6,<8.7",
-    #"torch >=2.0.0,<2.1.0",
-    "torch==2.1.0.dev20230703+cu121",
+    "torch >=2.1.0.dev,<2.2.0",
+    #"torch==2.1.0.dev20230731",
     "pybind11==2.6.2",
     "numpy",
 ]
@@ -42,8 +42,8 @@ readme = {file = "py/README.md", content-type = "text/markdown"}
 requires-python = ">=3.8"
 keywords = ["pytorch", "torch", "tensorrt", "trt", "ai", "artificial intelligence", "ml", "machine learning", "dl", "deep learning", "compiler", "dynamo", "torchscript", "inference"]
 dependencies = [
-    #"torch>=2.0.0,<2.1.0",
-    "torch==2.1.0.dev20230703+cu121",
+    "torch >=2.1.0.dev,<2.2.0",
+    #"torch==2.1.0.dev20230731",
     "tensorrt>=8.6,<8.7",
     "packaging>=23",
     "numpy",