Migrate CPU CI to GH

.azure_pipelines/olive-ci.yaml

@@ -33,12 +33,20 @@ variables:
   skipComponentGovernanceDetection: true
 
 jobs:
+- job: Olive_CI_NoOp
+  displayName: "Olive CI moved to GitHub Actions"
+  pool:
+    vmImage: "ubuntu-latest"
+  steps:
+  - script: echo "No Azure Pipelines jobs configured for this pipeline."
+    displayName: "No-op placeholder"
+
 # Linux unit tests
-- template: job_templates/olive-test-cpu-template.yaml
-  parameters:
-    name: Linux_CPU_CI_Unit_Test
-    pool: $(OLIVE_POOL_UBUNTU2004)
-    pytest_marker: 'not openvino and not amd'
+# - template: job_templates/olive-test-cpu-template.yaml
+#   parameters:
+#     name: Linux_CPU_CI_Unit_Test
+#     pool: $(OLIVE_POOL_UBUNTU2004)
+#     pytest_marker: 'not openvino and not amd'
 
 # Skip the GPU CI until the pool issue is resolved (not able to allocate GPU resources).
 # - template: job_templates/olive-test-linux-gpu-template.yaml
@@ -48,9 +56,9 @@ jobs:
 #     pytest_marker: 'not openvino and not amd'
 
 # Windows unit tests
-- template: job_templates/olive-test-cpu-template.yaml
-  parameters:
-    name: Windows_CPU_CI_Unit_Test
-    pool: $(OLIVE_POOL_WIN2019)
-    windows: True
-    pytest_marker: 'not openvino and not amd'
+# - template: job_templates/olive-test-cpu-template.yaml
+#   parameters:
+#     name: Windows_CPU_CI_Unit_Test
+#     pool: $(OLIVE_POOL_WIN2019)
+#     windows: True
+#     pytest_marker: 'not openvino and not amd'

.github/workflows/cpu-ci.yml

@@ -0,0 +1,128 @@
+name: Olive CPU CI
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - main
+  pull_request:
+    branches:
+      - main
+  issue_comment:
+    types:
+      - created
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event_name == 'issue_comment' && format('pr-{0}-{1}', github.event.issue.number, github.event.comment.id) || github.ref }}
+  cancel-in-progress: true
+
+permissions:
+  contents: read
+  pull-requests: read
+  issues: read
+
+jobs:
+  cpu_tests:
+    if: >-
+      github.event_name == 'push' ||
+      github.event_name == 'workflow_dispatch' ||
+      (github.event_name == 'pull_request' && github.event.pull_request.head.repo.fork == false) ||
+      (
+        github.event_name == 'issue_comment' &&
+        github.event.issue.pull_request != null &&
+        contains(github.event.comment.body, '/ci run') &&
+        (
+          github.event.comment.author_association == 'OWNER' ||
+          github.event.comment.author_association == 'MEMBER' ||
+          github.event.comment.author_association == 'COLLABORATOR'
+        )
+      )
+    name: ${{ matrix.name }} CPU unit tests
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - name: Linux
+            os: ubuntu-22.04
+            platform: linux
+          - name: Windows
+            os: windows-2022
+            platform: windows
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 300
+    env:
+      HF_HOME: ${{ github.workspace }}/.cache/huggingface
+      PYTHONIOENCODING: utf-8
+      PYTHONUTF8: "1"
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          ref: ${{ github.event_name == 'issue_comment' && format('refs/pull/{0}/merge', github.event.issue.number) || github.ref }}
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: pip
+          cache-dependency-path: |
+            requirements.txt
+            test/requirements-test.txt
+            test/requirements-test-cpu.txt
+
+      - name: Install PyTorch
+        run: python -m pip install torch
+
+      - name: Install Olive
+        run: python -m pip install .
+
+      - name: Install ONNX Runtime
+        run: python -m pip install onnxruntime
+
+      - name: Install pytest
+        run: python -m pip install pytest
+
+      - name: Install test dependencies
+        run: python -m pip install -r test/requirements-test-cpu.txt
+
+      - name: List installed packages
+        run: python -m pip list
+
+      - name: Log in to Hugging Face
+        env:
+          HF_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+        run: >-
+          python -c "import os; from huggingface_hub import login;
+          login(token=os.environ['HF_TOKEN'])"
+
+      - name: Create test output directory
+        run: python -c "from pathlib import Path; Path('logs').mkdir(exist_ok=True)"
+
+      - name: Run tests
+        run: >-
+          python -m coverage run --source=olive -m pytest -v -s -p no:warnings
+          --disable-warnings --log-cli-level=WARNING
+          --junitxml=logs/test-TestOlive-${{ matrix.platform }}.xml
+          -m "not openvino and not amd" test
+          --basetemp "${{ runner.temp }}/pytest"
+
+      - name: Show coverage report
+        if: always()
+        continue-on-error: true
+        run: python -m coverage report -m
+
+      - name: Generate coverage report
+        if: always()
+        continue-on-error: true
+        run: python -m coverage xml -o coverage-${{ matrix.platform }}.xml
+
+      - name: Upload test results
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: cpu-test-results-${{ matrix.platform }}
+          path: |
+            logs/*.xml
+            coverage-${{ matrix.platform }}.xml
+          if-no-files-found: warn
+          retention-days: 14

test/cli/test_cli_test_model_smoke.py

@@ -303,7 +303,7 @@ class TestCliTestModelSmoke(unittest.TestCase):
 
     def test_documented_test_model_smoke_flow(self):
         if self.workdir is None:
-            with tempfile.TemporaryDirectory() as temp_dir:
+            with tempfile.TemporaryDirectory(ignore_cleanup_errors=True) as temp_dir:
                 self._assert_smoke_flows(Path(temp_dir))
         else:
             workdir = Path(self.workdir)
@@ -352,7 +352,7 @@ def test_save_local_tiny_qwen2_5_vl_supports_image_processor(self):
             self.skipTest("Pillow is required for VLM processor smoke test.")
         from transformers import AutoProcessor
 
-        with tempfile.TemporaryDirectory() as temp_dir:
+        with tempfile.TemporaryDirectory(ignore_cleanup_errors=True) as temp_dir:
             model_path = Path(temp_dir) / "qwen2_5_vl"
             _save_local_tiny_qwen2_5_vl(model_path)
 
@@ -376,7 +376,7 @@ def test_save_local_tiny_qwen2_5_vl_supports_image_processor(self):
     def test_model_discrepancy(self):
         """Verify that OnnxDiscrepancyCheck runs successfully with the configured exporter."""
         if self.workdir is None:
-            with tempfile.TemporaryDirectory() as temp_dir:
+            with tempfile.TemporaryDirectory(ignore_cleanup_errors=True) as temp_dir:
                 self._assert_discrepancy(Path(temp_dir))
         else:
             workdir = Path(self.workdir)

test/cli/test_cli_whisper_smoke.py

@@ -97,7 +97,7 @@ class TestCliWhisperSmoke(unittest.TestCase):
     def test_whisper_capture_onnx_graph(self):
         """Verify that Whisper encoder and decoder are exported to ONNX successfully."""
         if self.workdir is None:
-            with tempfile.TemporaryDirectory() as temp_dir:
+            with tempfile.TemporaryDirectory(ignore_cleanup_errors=True) as temp_dir:
                 self._assert_whisper_export(Path(temp_dir))
         else:
             workdir = Path(self.workdir)

test/passes/onnx/test_mnb_to_qdq.py

@@ -150,10 +150,16 @@ def test_mnb_to_qdq(create_mnb_model, nodes_to_exclude, add_zero_point, use_sign
     # disable qdq to mnb fusion so we can test the output of the DQ nodes directly
     disabled_optimizers = ["QDQSelectorActionTransformer"]
     if is_symmetric and use_signed_int and not add_zero_point and use_transpose_op:
-        # there seems to be a bug in ORT graph optimization which changes the int4 DQ to uint8 DQ
-        with pytest.raises(Exception, match="uint8"):
-            onnxruntime.InferenceSession(str(qdq_model.model_path), disabled_optimizers=disabled_optimizers)
-        return
+        # Some ORT builds fail here because graph optimization changes signed sub-byte DQ to uint8 DQ.
+        try:
+            qdq_session = onnxruntime.InferenceSession(
+                str(qdq_model.model_path), disabled_optimizers=disabled_optimizers
+            )
+            qdq_session.disable_fallback()
+        except Exception as e:
+            if "uint8" in str(e):
+                return
+            raise
     else:
         qdq_session = onnxruntime.InferenceSession(str(qdq_model.model_path), disabled_optimizers=disabled_optimizers)
         qdq_session.disable_fallback()