Dockerize BH pipeline (#15523)

### Ticket Progress towards #14393 ### Problem description To run our pipeline on 22.04, we need to first dockerize the steps (else we need a whole new fleet of build runners). ### What's changed Run the Blackhole post-commit within Docker. Still using 20.04, but sets the stage for 22.04. ### Checklist - [ ] Post commit CI passes https://github.com/tenstorrent/tt-metal/actions/runs/12263532830 - [ ] Blackhole Post commit (if applicable) https://github.com/tenstorrent/tt-metal/actions/runs/12263536331 - [ ] Model regression CI testing passes (if applicable) - [ ] Device performance regression CI testing passes (if applicable) - [ ] New/Existing tests provide coverage for changes
tenstorrent · Dec 10, 2024 · 52dec89 · 52dec89
1 parent 208db3e
commit 52dec89
Show file tree

Hide file tree

Showing 5 changed files with 58 additions and 16 deletions.
diff --git a/.github/workflows/all-post-commit-workflows.yaml b/.github/workflows/all-post-commit-workflows.yaml
@@ -108,6 +108,7 @@ jobs:
         ]
     uses: ./.github/workflows/fast-dispatch-build-and-unit-tests.yaml
     with:
+      os: ubuntu-20.04
       arch: ${{ matrix.test-group.arch }}
       runner-label: ${{ matrix.test-group.runner-label }}
   # TTNN FD Unit tests

diff --git a/.github/workflows/blackhole-post-commit.yaml b/.github/workflows/blackhole-post-commit.yaml
@@ -29,6 +29,7 @@ jobs:
     uses: ./.github/workflows/build-artifact.yaml
     secrets: inherit
     with:
+      os: "ubuntu-20.04-amd64"
       arch: '["blackhole"]'
       build-docker: false
   build-wheels:
@@ -57,13 +58,15 @@ jobs:
       arch: blackhole
       runner-label: BH
       timeout: 30
+      os: "ubuntu-20.04"
   fd-unit-tests:
     needs: build-wheels
     uses: ./.github/workflows/fast-dispatch-build-and-unit-tests.yaml
     secrets: inherit
     with:
       arch: blackhole
       runner-label: BH
+      os: "ubuntu-20.04"
   # FD C++ Unit Tests
   cpp-unit-tests:
     needs: build-artifact
@@ -73,6 +76,7 @@ jobs:
       arch: blackhole
       runner-label: BH
       timeout: 60
+      os: "ubuntu-20.04"
 
 #   profiler-regression:
 #     needs: build-artifact-profiler

diff --git a/.github/workflows/build-and-unit-tests.yaml b/.github/workflows/build-and-unit-tests.yaml
@@ -13,6 +13,10 @@ on:
         required: false
         type: number
         default: 35
+      os:
+        required: false
+        type: string
+        default: "ubuntu-20.04"
   workflow_dispatch:
     inputs:
       arch:
@@ -34,6 +38,11 @@ on:
         required: false
         type: number
         default: 35
+      os:
+        required: false
+        type: string
+        default: "ubuntu-20.04"
+
 jobs:
   unit-tests-slow-dispatch:
     name: ${{ inputs.arch }} ${{ inputs.runner-label }}
@@ -42,24 +51,28 @@ jobs:
       - cloud-virtual-machine
       - in-service
     env:
-      TT_METAL_ENV: ${{ vars.TT_METAL_ENV }}
-      ARCH_NAME: ${{ inputs.arch}}
-      TT_METAL_SLOW_DISPATCH_MODE: 1
+      ARCH_NAME: ${{ inputs.arch }}
       LOGURU_LEVEL: INFO
-      LD_LIBRARY_PATH: ${{ github.workspace }}/build/lib
     steps:
       - uses: tenstorrent/tt-metal/.github/actions/checkout-with-submodule-lfs@main
       - uses: ./.github/actions/prepare-metal-run
         with:
           arch: ${{ inputs.arch }}
-      - name: Set up dynamic env vars for build
-        run: |
-          echo "TT_METAL_HOME=$(pwd)" >> $GITHUB_ENV
       - name: Run pre/post regression tests
         timeout-minutes: ${{ inputs.timeout }}
-        run: |
-          source ${{ github.workspace }}/python_env/bin/activate
-          ./tests/scripts/run_tests.sh --tt-arch $ARCH_NAME --pipeline-type post_commit --dispatch-mode slow
+        uses: ./.github/actions/docker-run
+        with:
+          docker_os_arch: tt-metalium/${{ inputs.os }}-amd64
+          docker_password: ${{ secrets.GITHUB_TOKEN }}
+          docker_opts: |
+            -e ARCH_NAME=${{ inputs.arch}}
+            -e TT_METAL_HOME=${{ github.workspace }}
+            -e TT_METAL_SLOW_DISPATCH_MODE=1
+            -e LD_LIBRARY_PATH=${{ github.workspace }}/build/lib
+          run_args: |
+            python3 -m pip install -r $(pwd)/tt_metal/python_env/requirements-dev.txt
+            pip install -e .
+            ./tests/scripts/run_tests.sh --tt-arch ${{ inputs.arch }} --pipeline-type post_commit --dispatch-mode slow
       - uses: ./.github/actions/slack-report
         if: ${{ failure() }}
         with:

diff --git a/.github/workflows/cpp-post-commit.yaml b/.github/workflows/cpp-post-commit.yaml
@@ -13,6 +13,10 @@ on:
         required: false
         type: number
         default: 80
+      os:
+        required: false
+        type: string
+        default: "ubuntu-20.04"
   workflow_dispatch:
     inputs:
       arch:
@@ -34,6 +38,10 @@ on:
         required: false
         type: number
         default: 60
+      os:
+        required: false
+        type: string
+        default: "ubuntu-20.04"
 
 jobs:
   models:
@@ -67,11 +75,18 @@ jobs:
           arch: ${{ inputs.arch }}
       - name: ${{ matrix.test-group.name }} tests
         timeout-minutes: ${{ inputs.timeout }}
-        run: |
-          source ${{ github.workspace }}/python_env/bin/activate
-          cd $TT_METAL_HOME
-          export PYTHONPATH=$TT_METAL_HOME
-          ${{ matrix.test-group.cmd }}
+        uses: ./.github/actions/docker-run
+        with:
+          docker_os_arch: tt-metalium/${{ inputs.os }}-amd64
+          docker_password: ${{ secrets.GITHUB_TOKEN }}
+          docker_opts: |
+            -e TT_METAL_HOME=${{ github.workspace }}
+            -e ARCH_NAME=${{ inputs.arch }}
+            -e LD_LIBRARY_PATH=${{ github.workspace }}/build/lib
+          run_args: |
+            python3 -m pip install -r $(pwd)/tt_metal/python_env/requirements-dev.txt
+            python3 -m pip install -e .
+            ${{ matrix.test-group.cmd }}
       - uses: ./.github/actions/slack-report
         if: ${{ failure() }}
         with:

diff --git a/.github/workflows/fast-dispatch-build-and-unit-tests.yaml b/.github/workflows/fast-dispatch-build-and-unit-tests.yaml
@@ -13,6 +13,10 @@ on:
         required: false
         type: number
         default: 45
+      os:
+        required: false
+        type: string
+        default: "ubuntu-20.04"
   workflow_dispatch:
     inputs:
       arch:
@@ -34,6 +38,10 @@ on:
         required: false
         type: number
         default: 45
+      os:
+        required: false
+        type: string
+        default: "ubuntu-20.04"
 
 jobs:
   fd-tests:
@@ -42,7 +50,7 @@ jobs:
       # so we try not to get hanging machines
       fail-fast: false
       matrix:
-        os: ["ubuntu-20.04"]
+        os: ["${{ inputs.os }}"]
         test-group: [
           {name: eager unit tests 1, cmd: pytest tests/tt_eager/python_api_testing/unit_testing/ -xvvv --splits 7 --group 1 },
           {name: eager unit tests 2, cmd: pytest tests/tt_eager/python_api_testing/unit_testing/ -xvvv --splits 7 --group 2 },
@@ -70,6 +78,7 @@ jobs:
         timeout-minutes: ${{ inputs.timeout }}
         uses: ./.github/actions/docker-run
         with:
+          docker_os_arch: tt-metalium/${{ inputs.os }}-amd64
           install_wheel: true
           docker_password: ${{ secrets.GITHUB_TOKEN }}
           run_args: |