[CI] Add initial github CI

.github/workflows/mlir-tensorrt-build-test.yml

@@ -0,0 +1,159 @@
+name: MLIR-TensorRT Build and Test
+
+on:
+  workflow_call:
+    inputs:
+      channel:
+        description: 'Channel, valid values are "nightly", "test", or "release"'
+        default: "test"
+        type: string
+      build-matrix:
+        description: 'Build matrix to utilize'
+        default: ""
+        type: string
+
+defaults:
+  run:
+    shell: bash
+
+jobs:
+  mlir-tensorrt-build-test:
+    name: build-test-${{ matrix.arch }}-cu${{ matrix.cuda }}-trt${{ matrix.trt }}
+    strategy:
+      fail-fast: false
+      matrix: ${{ fromJSON(inputs.build-matrix) }}
+    env:
+      # eg. CUDA_VERSION: 12.9.1 or 13.0.0
+      CUDA_VERSION: ${{ matrix.cuda }}
+      # eg. TENSORRT_VERSION: 10.12 or 10.13
+      TENSORRT_VERSION: ${{ matrix.trt }}
+      # eg. CHANNEL: nightly, test or release
+      CHANNEL: ${{ inputs.channel }}
+    runs-on: ${{ matrix.runner }}
+    container:
+      image: ${{ matrix.docker_image }}
+      options: '--gpus all'
+    steps:
+      # Checkout the repository
+      - name: Checkout TensorRT-Incubator
+        uses: actions/checkout@v4
+
+      # Create cache folders
+      - name: Create Cache Folders
+        run: |
+          set -euo pipefail
+          set -x
+          mkdir -p mlir-tensorrt/ccache
+          mkdir -p mlir-tensorrt/.cache.cpm
+
+      # Restore cache, if exists.
+      - name: Restore Cache
+        uses: actions/cache/restore@v4
+        with:
+          key: ${{ runner.os }}-mlir-tensorrt-cache-${{ hashFiles('mlir-tensorrt/**/*.cpp', 'mlir-tensorrt/**/*.h', 'mlir-tensorrt/build_tools/**/*') }}
+          restore-keys: |
+            ${{ runner.os }}-mlir-tensorrt-cache-
+          path: |
+            mlir-tensorrt/ccache
+            mlir-tensorrt/.cache.cpm/*
+            !mlir-tensorrt/.cache.cpm/tensorrt
+
+      # Build the project
+      - name: Build With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          # Build only, skip tests
+          SKIP_TESTS=1 \
+          ./build_tools/scripts/cicd_build_test.sh
+
+      # Save cache
+      - name: Save cache
+        uses: actions/cache/save@v4
+        with:
+          key: ${{ runner.os }}-mlir-tensorrt-cache-${{ hashFiles('mlir-tensorrt/**/*.cpp', 'mlir-tensorrt/**/*.h', 'mlir-tensorrt/build_tools/**/*') }}
+          path: |
+            mlir-tensorrt/ccache
+            mlir-tensorrt/.cache.cpm/*
+            !mlir-tensorrt/.cache.cpm/tensorrt
+
+      # Run tests
+      - name: Run Basic Tests With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        id: basic-tests
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          # run tests
+          ./build_tools/scripts/cicd_build_test.sh
+
+      # Run ASAN tests
+      - name: Run ASAN Tests With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        id: asan-tests
+        if: ${{ always() && inputs.channel == 'nightly' }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          CMAKE_PRESET=github-cicd-with-asan \
+          ./build_tools/scripts/cicd_build_test.sh
+
+      # Run Long tests
+      - name: Run Long Tests With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        id: long-tests
+        if: ${{ always() && inputs.channel == 'nightly' }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          CMAKE_PRESET=github-cicd-with-long-tests \
+          ./build_tools/scripts/cicd_build_test.sh
+
+      # Run NCCL Long tests
+      - name: Run NCCL Long Tests With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        id: nccl-long-tests
+        if: ${{ always() && inputs.channel == 'nightly' }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          CMAKE_PRESET=github-cicd-with-nccl-long-tests \
+          ./build_tools/scripts/cicd_build_test.sh
+
+      # Build wheels
+      - name: Build Wheels With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        id: build-wheels
+        if: ${{ always() && inputs.channel == 'nightly' }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          ./build_tools/scripts/cicd_build_wheels.sh
+
+      # Smoke test the wheels built
+      - name: Smoke Test Wheels
+        id: smoke-test-wheels
+        if: ${{ inputs.channel == 'nightly' }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          for whl in .wheels/*.whl; do
+            uv pip install "$whl"
+          done
+          # TODO: add smoke test for the wheel
+
+      # Build distribution tar.gz with CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+      - name: Build Distribution Tar.gz With CUDA:${{ matrix.cuda }} + TensorRT:${{ matrix.trt }}
+        id: build-distribution
+        if: ${{ always() && inputs.channel == 'nightly' }}
+        run: |
+          set -euo pipefail
+          set -x
+          cd mlir-tensorrt
+          ./build_tools/scripts/cicd_build_distribution.sh
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-mlir-tensorrt-${{ inputs.channel }}
+  cancel-in-progress: true

.github/workflows/mlir-tensorrt-ci.yml

@@ -1,123 +1,97 @@
 name: MLIR-TensorRT CI
 
 on:
-  pull_request:
-    branches:
-      - main
-    types: [synchronize, opened, reopened, ready_for_review]
-    paths: ["mlir-tensorrt/**"]
   push:
     branches:
-      - main
-    paths: ["mlir-tensorrt/**"]
+      - "pull-request/[0-9]+"
+    tags:
+      # release tag example: v0.4.2
+      # release candidate tag example: v0.4.2-rc1
+      - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
+  workflow_dispatch:
+  schedule:
+    - cron: '0 0 * * *'  # Runs at 00:00 UTC every day (minute hour day-of-month month-of-year day-of-week)
 
 env:
-  DEFAULT_IMAGE: ghcr.io/nvidia/tensorrt-incubator/mlir-tensorrt:cuda12.5-ubuntu-llvm17
+  DEFAULT_IMAGE: ghcr.io/nvidia/tensorrt-incubator/mlir-tensorrt:cuda12.9-ubuntu-llvm17
   REGISTRY: ghcr.io
 
-jobs:
-  mlir-tensorrt-test-pr:
-    if: github.event_name != 'pull_request' || github.event.pull_request.draft == false
-    runs-on: tripy-self-hosted
+defaults:
+  run:
+    shell: bash
 
+jobs:
+  format-check:
+    name: Lint Check
+    runs-on: ubuntu-latest
+    container:
+      image: ghcr.io/nvidia/tensorrt-incubator/mlir-tensorrt:cuda12.9-ubuntu-llvm17
     steps:
-      # Value of `github.workspace` is /home/runner/work/{repo_name}/{repo-name}
-      # i.e. /home/runner/work/TensorRT-Incubator/TensorRT-Incubator in our case.
-      # After this action, repo is cloned inside above path.
-      - uses: actions/checkout@v4
+      - name: checkout TensorRT-Incubator
+        uses: actions/checkout@v4
         with:
           fetch-depth: 5
-
-      # Run initial format check
       - name: Run python format and clang check
-        uses: addnab/docker-run-action@v3
-        if: ${{ github.event_name == 'pull_request' }}
-        with:
-          image: ${{ env.DEFAULT_IMAGE }}
-          options: -v ${{ github.workspace }}:/tensorrt-incubator
-          registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-          # This step does two things
-          # 1. Check if Python files follow black format
-          # 2. Check if C++ files follow clang format
-          # NOTE: We are placed at the root directory ('/') inside the container.
-          run: |
-            cd tensorrt-incubator
-            git config --global --add safe.directory /tensorrt-incubator
-            cat > run_format_check.sh <<EOF
-            #!/bin/bash
-            set -e
-            python3 -m black --check --extend-exclude='.*\.pyi' mlir-tensorrt/compiler/
-            python3 -m black --check --extend-exclude='.*\.pyi' mlir-tensorrt/integrations/python/
-            git clang-format HEAD~1 --diff
-            EOF
+        shell: bash
+        run: |
+          set -euo pipefail
+          set -x
 
-            bash run_format_check.sh
+          uv venv .venv || { echo "Failed to create virtual environment"; exit 1; }
+          source .venv/bin/activate || { echo "Failed to activate virtual environment"; exit 1; }
+          uv pip install black
+          uv run black --check --extend-exclude='.*\.pyi' \
+          mlir-tensorrt/compiler/ mlir-tensorrt/integrations/python/
 
-      # Create cache folders
-      - name: Create cache folder
-        run: |
-          mkdir -p ${{ github.workspace }}/ccache
-          mkdir -p ${{ github.workspace }}/.cache.cpm
+          git config --global --add safe.directory $(pwd)
+          uv run git clang-format HEAD~1 --diff
 
-      # Restore cache, if exists.
-      - name: Restore cache
-        id: restore-cache
-        uses: actions/cache/restore@v4
-        with:
-          key: ${{ runner.os }}-mlir-tensorrt-cache-${{ hashFiles('mlir-tensorrt/**/*.cpp', 'mlir-tensorrt/**/*.h', 'mlir-tensorrt/build_tools/**/*') }}
-          restore-keys: |
-            ${{ runner.os }}-mlir-tensorrt-cache-
-          path: |
-            ${{ github.workspace }}/ccache
-            ${{ github.workspace }}/.cache.cpm/*
-            !${{ github.workspace }}/.cache.cpm/tensorrt
+  generate-matrix:
+    name: Generate Build Matrix
+    runs-on: ubuntu-latest
+    needs: format-check
+    outputs:
+      matrix: ${{ steps.generate.outputs.matrix }}
+      channel: ${{ steps.generate.outputs.channel }}
+    steps:
+      - name: Checkout TensorRT-Incubator
+        uses: actions/checkout@v4
 
-      # TensorRT 10 tests
-      - name: TensorRT 10 build
-        uses: addnab/docker-run-action@v3
-        with:
-          image: ${{ env.DEFAULT_IMAGE }}
-          options: -v ${{ github.workspace }}/mlir-tensorrt:/mlir-tensorrt -v ${{ github.workspace }}/ccache:/ccache -v ${{ github.workspace }}/.cache.cpm:/.cache.cpm --gpus all
-          registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-          run: |
-            cd mlir-tensorrt
-            ./build_tools/scripts/cicd_build.sh --build_only
+      - name: Generate Build Matrix
+        id: generate
+        run: |
+          set -euo pipefail
+          set -x
+          CHANNEL="test"
+          if [ "${GITHUB_EVENT_NAME:-}" = "schedule" ]; then
+            CHANNEL="nightly"
+          elif [ "${GITHUB_REF_TYPE:-}" = "tag" ]; then
+            CHANNEL="release"
+          fi
+          MATRIX_BLOB="$(python3 ./.github/workflows/mlir-tensorrt/generate-matrix.py --channel "${CHANNEL}")"
+          echo "${MATRIX_BLOB}"
+          echo "${CHANNEL}"
+          echo "matrix=${MATRIX_BLOB}" >> "${GITHUB_OUTPUT}"
+          echo "channel=${CHANNEL}" >> "${GITHUB_OUTPUT}"
 
-      - name: Save cache
-        id: save-cache
-        uses: actions/cache/save@v4
-        with:
-          key: ${{ runner.os }}-mlir-tensorrt-cache-${{ hashFiles('mlir-tensorrt/**/*.cpp', 'mlir-tensorrt/**/*.h', 'mlir-tensorrt/build_tools/**/*') }}
-          path: |
-            ${{ github.workspace }}/ccache
-            ${{ github.workspace }}/.cache.cpm/*
-            !${{ github.workspace }}/.cache.cpm/tensorrt
+  mlir-tensorrt-build-test:
+    name: Build and Test
+    needs: generate-matrix
+    if: ${{ needs.generate-matrix.outputs.channel != 'release' }}
+    uses: ./.github/workflows/mlir-tensorrt-build-test.yml
+    with:
+      build-matrix: ${{ needs.generate-matrix.outputs.matrix }}
+      channel: ${{ needs.generate-matrix.outputs.channel }}
 
-      - name: TensorRT 10 test
-        uses: addnab/docker-run-action@v3
-        with:
-          image: ${{ env.DEFAULT_IMAGE }}
-          options: -v ${{ github.workspace }}/mlir-tensorrt:/mlir-tensorrt -v ${{ github.workspace }}/ccache:/ccache -v ${{ github.workspace }}/.cache.cpm:/.cache.cpm --gpus all
-          registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-          run: |
-            cd mlir-tensorrt
-            ./build_tools/scripts/cicd_build.sh
+  mlir-tensorrt-release:
+    name: Release
+    needs: generate-matrix
+    if: ${{ needs.generate-matrix.outputs.channel == 'release' }}
+    uses: ./.github/workflows/mlir-tensorrt-release.yml
+    with:
+      build-matrix: ${{ needs.generate-matrix.outputs.matrix }}
+      channel: release
 
-      # TensorRT 10 & ASAN
-      - name: TensorRT 10 ASAN test
-        uses: addnab/docker-run-action@v3
-        with:
-          image: ${{ env.DEFAULT_IMAGE }}
-          options: -v ${{ github.workspace }}/mlir-tensorrt:/mlir-tensorrt -v ${{ github.workspace }}/ccache:/ccache -v ${{ github.workspace }}/.cache.cpm:/.cache.cpm --gpus all
-          registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-          run: |
-            cd mlir-tensorrt
-            ENABLE_ASAN=ON ./build_tools/scripts/cicd_build.sh
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-mlir-tensorrt
+  cancel-in-progress: true