Merge remote-tracking branch 'origin' into karenc/distributed_blocks

Author: karen-sy

.github/actions/pytest/action.yml

@@ -54,7 +54,16 @@ runs:
         # Run pytest with detailed output and JUnit XML
         set +e  # Don't exit on test failures
 
-        docker run --runtime=nvidia --gpus all -w /workspace \
+        # Detect GPU availability and conditionally add GPU flags
+        GPU_FLAGS=""
+        if command -v nvidia-smi &> /dev/null && nvidia-smi &> /dev/null; then
+          echo "GPU detected, enabling GPU runtime"
+          GPU_FLAGS="--runtime=nvidia --gpus all"
+        else
+          echo "No GPU detected, running in CPU-only mode"
+        fi
+
+        docker run ${GPU_FLAGS} --rm -w /workspace \
           --cpus=${NUM_CPUS} \
           --network host \
           --name ${{ env.CONTAINER_ID }}_pytest \

.github/workflows/container-validation-backends.yml

@@ -318,7 +318,7 @@ jobs:
         uses: ./.github/actions/pytest
         with:
           image_tag: ${{ steps.build-image.outputs.image_tag }}
-          pytest_marks: "unit and trtllm_marker and gpu_1"
+          pytest_marks: "unit and trtllm and gpu_1"
           framework: "trtllm"
           test_type: "unit"
           platform_arch: ${{ matrix.platform.arch }}
@@ -327,7 +327,7 @@ jobs:
         uses: ./.github/actions/pytest
         with:
           image_tag: ${{ steps.build-image.outputs.image_tag }}
-          pytest_marks: "e2e and trtllm_marker and gpu_1 and not slow"
+          pytest_marks: "e2e and trtllm and gpu_1 and not slow"
           framework: "trtllm"
           test_type: "e2e, gpu_1"
           platform_arch: ${{ matrix.platform.arch }}

.github/workflows/container-validation-dynamo.yml

@@ -65,7 +65,7 @@ jobs:
           docker compose down
       - name: Run pytest (parallel tests with xdist)
         env:
-          PYTEST_MARKS: "pre_merge and parallel"
+          PYTEST_MARKS: "pre_merge and parallel and not (vllm or trtllm or sglang or k8s)"
         run: |
           docker run -w /workspace \
             --name ${{ env.CONTAINER_ID }}_pytest_parallel \
@@ -77,7 +77,7 @@ jobs:
           docker cp ${{ env.CONTAINER_ID }}_pytest_parallel:/workspace/${{ env.PYTEST_PARALLEL_XML_FILE }} . || echo "No parallel test report found"
       - name: Run pytest (sequential tests)
         env:
-          PYTEST_MARKS: "(pre_merge and not parallel) or mypy"
+          PYTEST_MARKS: "(pre_merge and not parallel and not (vllm or trtllm or sglang or k8s)) or mypy"
         run: |
           docker run -w /workspace \
             --name ${{ env.CONTAINER_ID }}_pytest \

components/src/dynamo/sglang/tests/test_sglang_prometheus_utils.py

@@ -11,6 +11,9 @@
 
 pytestmark = [
     pytest.mark.unit,
+    pytest.mark.sglang,
+    pytest.mark.gpu_0,
+    pytest.mark.post_merge,
 ]
 
 

components/src/dynamo/sglang/tests/test_sglang_unit.py

@@ -19,13 +19,14 @@
 JINJA_TEMPLATE_PATH = str(
     REPO_ROOT / "tests" / "serve" / "fixtures" / "custom_template.jinja"
 )
+
 pytestmark = [
     pytest.mark.unit,
     pytest.mark.sglang,
     pytest.mark.gpu_1,
+    pytest.mark.pre_merge,
 ]
 
-
 # Create SGLang-specific CLI args fixture
 # This will use monkeypatch to write to argv
 mock_sglang_cli = make_cli_args_fixture("dynamo.sglang")

components/src/dynamo/trtllm/tests/test_trtllm_autodeploy.py

@@ -14,7 +14,7 @@
 
 pytestmark = [
     pytest.mark.unit,
-    pytest.mark.trtllm_marker,
+    pytest.mark.trtllm,
     # NOTE: these tests do not actually require a GPU, but the workflow validation
     # `.github/workflows/container-validation-backends.yml` does not make use of
     # the `gpu_0` marker.

components/src/dynamo/trtllm/tests/test_trtllm_main_init.py

@@ -10,7 +10,12 @@
 import pytest
 
 # Mark all tests in this module to run only in TensorRT-LLM container
-pytestmark = pytest.mark.trtllm
+pytestmark = [
+    pytest.mark.unit,
+    pytest.mark.trtllm,
+    pytest.mark.gpu_1,
+    pytest.mark.pre_merge,
+]
 
 
 def test_tensorrt_llm_metrics_collector_import():

components/src/dynamo/trtllm/tests/test_trtllm_prometheus_utils.py

@@ -11,6 +11,9 @@
 
 pytestmark = [
     pytest.mark.unit,
+    pytest.mark.trtllm,
+    pytest.mark.gpu_0,
+    pytest.mark.post_merge,
 ]
 
 

components/src/dynamo/trtllm/tests/test_trtllm_unit.py

@@ -21,7 +21,7 @@
 
 pytestmark = [
     pytest.mark.unit,
-    pytest.mark.trtllm_marker,
+    pytest.mark.trtllm,
     pytest.mark.gpu_1,
 ]
 

components/src/dynamo/vllm/tests/test_vllm_prometheus_utils.py

@@ -11,6 +11,9 @@
 
 pytestmark = [
     pytest.mark.unit,
+    pytest.mark.vllm,
+    pytest.mark.gpu_0,
+    pytest.mark.post_merge,
 ]