[ViPPET] Simplify encoded video output configuration (auto-select CPU/GPU encoder)

tools/visual-pipeline-and-platform-evaluation-tool/vippet/api/api_schemas.py

@@ -528,63 +528,24 @@ class PipelineRequestOptimize(BaseModel):
     parameters: Optional[Dict[str, Any]]
 
 
-class EncoderDeviceConfig(BaseModel):
-    """
-    Encoder device configuration used in video output settings.
-
-    Attributes:
-        device_name: Name of the encoder device (for example ``"GPU"``).
-        gpu_id: Optional GPU index when applicable.
-
-    Example:
-        .. code-block:: json
-
-            {
-              "device_name": "GPU",
-              "gpu_id": 0
-            }
-    """
-
-    device_name: str = Field(
-        default="GPU",
-        description="Name of the encoder device (e.g., 'GPU', 'CPU', 'NPU')",
-        examples=["GPU", "CPU", "NPU"],
-    )
-    gpu_id: Optional[int] = Field(
-        default=None,
-        description="GPU device index (only applicable when device_name indicates a GPU)",
-        examples=[0, 1],
-    )
-
-
 class VideoOutputConfig(BaseModel):
     """
     Generic configuration of optional encoded video output.
 
     Attributes:
         enabled: Flag to enable or disable video output generation.
-        encoder_device: EncoderDeviceConfig used when video output is enabled.
 
     Example:
         .. code-block:: json
 
             {
-              "enabled": false,
-              "encoder_device": {
-                "device_name": "GPU",
-                "gpu_id": 0
-              }
+              "enabled": false
             }
     """
 
     enabled: bool = Field(
         default=False, description="Flag to enable or disable video output generation."
     )
-    encoder_device: EncoderDeviceConfig = Field(
-        default=EncoderDeviceConfig(device_name="GPU", gpu_id=0),
-        description="Encoder device configuration (only applicable when video output is enabled).",
-        examples=[{"device_name": "GPU", "gpu_id": 0}],
-    )
 
 
 class PerformanceTestSpec(BaseModel):
@@ -609,12 +570,9 @@ class PerformanceTestSpec(BaseModel):
     video_output: VideoOutputConfig = Field(
         default=VideoOutputConfig(
             enabled=False,
-            encoder_device=EncoderDeviceConfig(device_name="GPU", gpu_id=0),
         ),
         description="Video output configuration.",
-        examples=[
-            {"enabled": False, "encoder_device": {"device_name": "GPU", "gpu_id": 0}}
-        ],
+        examples=[{"enabled": False}],
     )
 
 
@@ -646,12 +604,9 @@ class DensityTestSpec(BaseModel):
     video_output: VideoOutputConfig = Field(
         default=VideoOutputConfig(
             enabled=False,
-            encoder_device=EncoderDeviceConfig(device_name="GPU", gpu_id=0),
         ),
         description="Video output configuration.",
-        examples=[
-            {"enabled": False, "encoder_device": {"device_name": "GPU", "gpu_id": 0}}
-        ],
+        examples=[{"enabled": False}],
     )
 
 

tools/visual-pipeline-and-platform-evaluation-tool/vippet/api/routes/tests.py

@@ -47,7 +47,7 @@ def run_performance_test(body: schemas.PerformanceTestSpec):
             * pipeline_performance_specs – list of pipelines and number of
               streams per pipeline.
             * video_output – configuration for optional encoded video output
-              (enabled flag and encoder_device).
+              (enabled flag).
 
     Returns:
         202 Accepted:
@@ -78,8 +78,7 @@ def run_performance_test(body: schemas.PerformanceTestSpec):
                 {"id": "pipeline-b7c2e114", "streams": 4}
               ],
               "video_output": {
-                "enabled": false,
-                "encoder_device": {"device_name": "GPU", "gpu_id": 0}
+                "enabled": false
               }
             }
 
@@ -186,8 +185,7 @@ def run_density_test(body: schemas.DensityTestSpec):
                 {"id": "pipeline-b7c2e114", "stream_rate": 50}
               ],
               "video_output": {
-                "enabled": false,
-                "encoder_device": {"device_name": "GPU", "gpu_id": 0}
+                "enabled": false
               }
             }
 

tools/visual-pipeline-and-platform-evaluation-tool/vippet/graph.py

@@ -8,6 +8,7 @@
 from utils import generate_unique_filename
 from videos import get_videos_manager, OUTPUT_VIDEO_DIR
 from models import get_supported_models_manager
+from video_encoder import ENCODER_DEVICE_CPU, ENCODER_DEVICE_GPU
 from resources import (
     get_labels_manager,
     get_scripts_manager,
@@ -367,6 +368,33 @@ def get_input_video_filenames(self) -> list[str]:
 
         return input_filenames
 
+    def get_recommended_encoder_device(self) -> str:
+        """
+        Iterate backwards through nodes to find the last video/x-raw caps node
+        and return the recommended encoder device based on memory type.
+
+        Note: NPU variants are not considered because NPUs do not provide dedicated
+        memory accessible for GStreamer pipeline buffering; they operate exclusively
+        on system or shared memory.
+
+        Returns:
+            str: ENCODER_DEVICE_GPU if video/x-raw(memory:VAMemory) is detected,
+                 ENCODER_DEVICE_CPU for standard video/x-raw or when no video/x-raw
+                 caps node exists in the pipeline.
+        """
+        for node in reversed(self.nodes):
+            if node.data.get(NODE_KIND_KEY) != NODE_KIND_CAPS:
+                continue
+
+            if not node.type.startswith("video/x-raw"):
+                continue
+
+            if "memory:VAMemory" in node.type:
+                return ENCODER_DEVICE_GPU
+            return ENCODER_DEVICE_CPU
+
+        return ENCODER_DEVICE_CPU
+
 
 @dataclass
 class _Token:

tools/visual-pipeline-and-platform-evaluation-tool/vippet/managers/pipeline_manager.py

@@ -299,11 +299,14 @@ def build_pipeline_command(
 
                 # Handle final video output if enabled
                 if video_config.enabled and stream_index == 0:
+                    # Get recommended encoder device from the graph
+                    encoder_device = graph.get_recommended_encoder_device()
+                    # Replace fakesink with actual video output element
                     unique_pipeline_str, generated_paths = (
                         self.video_encoder.replace_fakesink_with_video_output(
                             pipeline.id,
                             unique_pipeline_str,
-                            video_config.encoder_device,
+                            encoder_device,
                             input_video_filenames,
                         )
                     )

tools/visual-pipeline-and-platform-evaluation-tool/vippet/tests/api_tests/tests_test.py

@@ -200,8 +200,6 @@ def test_run_performance_test_with_gpu_encoder(self, mock_test_manager):
 
         # Verify video output configuration
         self.assertTrue(call_args.video_output.enabled)
-        self.assertEqual(call_args.video_output.encoder_device.device_name, "GPU")
-        self.assertEqual(call_args.video_output.encoder_device.gpu_id, 0)
 
     # ------------------------------------------------------------------
     # /tests/density

tools/visual-pipeline-and-platform-evaluation-tool/vippet/tests/graph_test.py

@@ -3,6 +3,7 @@
 from dataclasses import dataclass
 from unittest.mock import MagicMock, patch
 from graph import Graph, Node, Edge
+from video_encoder import ENCODER_DEVICE_GPU, ENCODER_DEVICE_CPU
 
 mock_models_manager = MagicMock()
 mock_videos_manager = MagicMock()
@@ -1505,5 +1506,133 @@ def test_empty_graph_raises_error(self):
         self.assertIn("Empty graph", str(cm.exception))
 
 
+class TestGetRecommendedEncoderDevice(unittest.TestCase):
+    """Test cases for Graph.get_recommended_encoder_device method."""
+
+    def test_gpu_encoder_for_va_memory_caps(self):
+        """Test that GPU encoder is recommended when video/x-raw(memory:VAMemory) is found."""
+        graph = Graph(
+            nodes=[
+                Node(id="0", type="filesrc", data={"location": "test.mp4"}),
+                Node(id="1", type="decodebin3", data={}),
+                Node(
+                    id="2",
+                    type="video/x-raw(memory:VAMemory)",
+                    data={"__node_kind": "caps"},
+                ),
+                Node(id="3", type="fakesink", data={}),
+            ],
+            edges=[
+                Edge(id="0", source="0", target="1"),
+                Edge(id="1", source="1", target="2"),
+                Edge(id="2", source="2", target="3"),
+            ],
+        )
+
+        self.assertEqual(graph.get_recommended_encoder_device(), ENCODER_DEVICE_GPU)
+
+    def test_cpu_encoder_for_standard_video_raw(self):
+        """Test that CPU encoder is recommended for standard video/x-raw caps."""
+        graph = Graph(
+            nodes=[
+                Node(id="0", type="filesrc", data={"location": "test.mp4"}),
+                Node(id="1", type="decodebin3", data={}),
+                Node(
+                    id="2",
+                    type="video/x-raw",
+                    data={"__node_kind": "caps", "width": "640", "height": "480"},
+                ),
+                Node(id="3", type="fakesink", data={}),
+            ],
+            edges=[
+                Edge(id="0", source="0", target="1"),
+                Edge(id="1", source="1", target="2"),
+                Edge(id="2", source="2", target="3"),
+            ],
+        )
+
+        self.assertEqual(graph.get_recommended_encoder_device(), ENCODER_DEVICE_CPU)
+
+    def test_cpu_encoder_when_no_video_raw_caps(self):
+        """Test that CPU encoder is recommended when no video/x-raw caps exist."""
+        graph = Graph(
+            nodes=[
+                Node(id="0", type="filesrc", data={"location": "test.mp4"}),
+                Node(id="1", type="decodebin3", data={}),
+                Node(id="2", type="queue", data={}),
+                Node(id="3", type="fakesink", data={}),
+            ],
+            edges=[
+                Edge(id="0", source="0", target="1"),
+                Edge(id="1", source="1", target="2"),
+                Edge(id="2", source="2", target="3"),
+            ],
+        )
+
+        self.assertEqual(graph.get_recommended_encoder_device(), ENCODER_DEVICE_CPU)
+
+    def test_uses_last_video_raw_caps_when_multiple_exist(self):
+        """Test that the method uses the last video/x-raw caps in the pipeline."""
+        graph = Graph(
+            nodes=[
+                Node(id="0", type="filesrc", data={"location": "test.mp4"}),
+                Node(
+                    id="1",
+                    type="video/x-raw",
+                    data={"__node_kind": "caps", "width": "640"},
+                ),
+                Node(id="2", type="queue", data={}),
+                Node(
+                    id="3",
+                    type="video/x-raw(memory:VAMemory)",
+                    data={"__node_kind": "caps"},
+                ),
+                Node(id="4", type="fakesink", data={}),
+            ],
+            edges=[
+                Edge(id="0", source="0", target="1"),
+                Edge(id="1", source="1", target="2"),
+                Edge(id="2", source="2", target="3"),
+                Edge(id="3", source="3", target="4"),
+            ],
+        )
+
+        # Should return GPU because the last video/x-raw has VAMemory
+        self.assertEqual(graph.get_recommended_encoder_device(), ENCODER_DEVICE_GPU)
+
+    def test_iterates_backwards_through_nodes(self):
+        """Test that the method iterates backwards (uses last occurrence, not first)."""
+        graph = Graph(
+            nodes=[
+                Node(id="0", type="filesrc", data={"location": "test.mp4"}),
+                Node(
+                    id="1",
+                    type="video/x-raw(memory:VAMemory)",
+                    data={"__node_kind": "caps"},
+                ),
+                Node(id="2", type="queue", data={}),
+                Node(
+                    id="3", type="video/x-raw", data={"__node_kind": "caps"}
+                ),  # Last one, no VAMemory
+                Node(id="4", type="fakesink", data={}),
+            ],
+            edges=[
+                Edge(id="0", source="0", target="1"),
+                Edge(id="1", source="1", target="2"),
+                Edge(id="2", source="2", target="3"),
+                Edge(id="3", source="3", target="4"),
+            ],
+        )
+
+        # Should return CPU because iterating backwards finds node 3 first (no VAMemory)
+        self.assertEqual(graph.get_recommended_encoder_device(), ENCODER_DEVICE_CPU)
+
+    def test_empty_graph(self):
+        """Test that CPU encoder is recommended for an empty graph."""
+        graph = Graph(nodes=[], edges=[])
+
+        self.assertEqual(graph.get_recommended_encoder_device(), ENCODER_DEVICE_CPU)
+
+
 if __name__ == "__main__":
     unittest.main()

tools/visual-pipeline-and-platform-evaluation-tool/vippet/tests/managers_tests/pipeline_manager_test.py

@@ -5,7 +5,6 @@
 from api.api_schemas import (
     Node,
     Edge,
-    EncoderDeviceConfig,
     PipelineType,
     PipelineSource,
     PipelineGraph,
@@ -458,7 +457,6 @@ def test_build_pipeline_command_with_video_output_enabled(self):
         pipeline_performance_specs = [PipelinePerformanceSpec(id=added.id, streams=1)]
         video_config = VideoOutputConfig(
             enabled=True,
-            encoder_device=EncoderDeviceConfig(device_name="CPU", gpu_id=None),
         )
 
         command, output_paths = manager.build_pipeline_command(
@@ -497,7 +495,6 @@ def test_build_pipeline_command_with_gpu_encoder(self):
         pipeline_performance_specs = [PipelinePerformanceSpec(id=added.id, streams=2)]
         video_config = VideoOutputConfig(
             enabled=True,
-            encoder_device=EncoderDeviceConfig(device_name="GPU", gpu_id=0),
         )
 
         command, output_paths = manager.build_pipeline_command(
@@ -544,7 +541,6 @@ def test_build_pipeline_command_video_output_multiple_pipelines(self):
         ]
         video_config = VideoOutputConfig(
             enabled=True,
-            encoder_device=EncoderDeviceConfig(device_name="CPU", gpu_id=None),
         )
 
         command, output_paths = manager.build_pipeline_command(