Yolo26 example

Author: daniil-lyakhov

.ci/scripts/test_yolo12.sh .ci/scripts/test_yolo26.sh.ci/scripts/test_yolo12.sh renamed to .ci/scripts/test_yolo26.sh

@@ -76,7 +76,7 @@ if [[ "${MODE}" =~ .*openvino.* ]]; then
   TARGET_LIBS="$TARGET_LIBS openvino_backend "
 
   git clone https://github.com/openvinotoolkit/openvino.git
-  cd openvino && git b16b776ac119dafda51f69a80f1e6b7376d02c3b
+  cd openvino && git checkout c694fbc2b6d
   git submodule update --init --recursive
   sudo ./install_build_dependencies.sh
   mkdir build && cd build
@@ -104,8 +104,8 @@ fi
 which "${PYTHON_EXECUTABLE}"
 
 
-DIR="examples/models/yolo12"
-$PYTHON_EXECUTABLE -m pip install -r ${DIR}/requirements.txt
+DIR="examples/models/yolo26"
+$PYTHON_EXECUTABLE -m pip install --upgrade-strategy only-if-needed -r ${DIR}/requirements.txt
 
 cmake_install_executorch_libraries() {
     rm -rf cmake-out
@@ -146,7 +146,7 @@ cmake_install_executorch_libraries() {
 }
 
 cmake_build_demo() {
-    echo "Building yolo12 runner"
+    echo "Building yolo26 runner"
     retry cmake \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \
         -DUSE_OPENVINO_BACKEND="$OPENVINO" \
@@ -181,17 +181,17 @@ EXPORT_ARGS="--model_name=${MODEL_NAME} --backend=${MODE}"
 # Add dynamically linked library location
 cmake_install_executorch_libraries
 
-$PYTHON_EXECUTABLE -m examples.models.yolo12.export_and_validate ${EXPORT_ARGS}
+$PYTHON_EXECUTABLE -m examples.models.yolo26.export_and_validate ${EXPORT_ARGS}
 
 
 RUNTIME_ARGS="--model_path=${EXPORTED_MODEL_NAME} --input_path=${VIDEO_PATH}"
 # Check build tool.
 cmake_build_demo
-# Run yolo12 runner
+# Run yolo26 runner
 NOW=$(date +"%H:%M:%S")
-echo "Starting to run yolo12 runner at ${NOW}"
+echo "Starting to run yolo26 runner at ${NOW}"
 # shellcheck source=/dev/null
-cmake-out/examples/models/yolo12/Yolo12DetectionDemo ${RUNTIME_ARGS} > result.txt
+cmake-out/examples/models/yolo26/Yolo26DetectionDemo ${RUNTIME_ARGS} > result.txt
 NOW=$(date +"%H:%M:%S")
 echo "Finished at ${NOW}"
 

README.md

@@ -204,7 +204,7 @@ ExecuTorch powers on-device AI at scale across Meta's family of apps, VR/AR devi
 
 **Multimodal:** [Llava](examples/models/llava/README.md) (vision-language), [Voxtral](examples/models/voxtral/README.md) (audio-language), [Gemma](examples/models/gemma3) (vision-language)
 
-**Vision/Speech:** [MobileNetV2](https://github.com/meta-pytorch/executorch-examples/tree/main/mv2), [DeepLabV3](https://github.com/meta-pytorch/executorch-examples/tree/main/dl3), [Whisper](examples/models/whisper/README.md) <!-- @lint-ignore -->
+**Vision/Speech:** [MobileNetV2](https://github.com/meta-pytorch/executorch-examples/tree/main/mv2), [DeepLabV3](https://github.com/meta-pytorch/executorch-examples/tree/main/dl3), [YOLO26](examples/models/yolo26/README.md), [Whisper](examples/models/whisper/README.md) <!-- @lint-ignore -->
 
 **Resources:** [`examples/`](examples/) directory • [executorch-examples](https://github.com/meta-pytorch/executorch-examples) out-of-tree demos • [Optimum-ExecuTorch](https://github.com/huggingface/optimum-executorch) for HuggingFace models • [Unsloth](https://docs.unsloth.ai/new/deploy-llms-phone) for fine-tuned LLM deployment <!-- @lint-ignore -->
 

backends/openvino/quantizer/quantizer.py

@@ -192,6 +192,8 @@ def set_ignored_scope(
         :param validate: If set to True, then a RuntimeError will be raised if any ignored scope does not match
           in the model graph.
         """
+        if subgraphs:
+            subgraphs = [nncf.Subgraph(inputs=subgraph[0], outputs=subgraph[1]) for subgraph in subgraphs]
         self._algo.set_ignored_scope(
             nncf.IgnoredScope(
                 names=names or [],

examples/models/yolo12/requirements.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.5)
 
-project(Yolo12DetectionDemo VERSION 0.1)
+project(Yolo26DetectionDemo VERSION 0.1)
 
 option(USE_OPENVINO_BACKEND "Build the tutorial with the OPENVINO backend" OFF)
 option(USE_XNNPACK_BACKEND "Build the tutorial with the XNNPACK backend" OFF)
@@ -63,14 +63,14 @@ set(PROJECT_SOURCES
     ${EXECUTORCH_ROOT}/extension/runner_util/inputs_portable.cpp
 )
 
-add_executable(Yolo12DetectionDemo ${PROJECT_SOURCES})
+add_executable(Yolo26DetectionDemo ${PROJECT_SOURCES})
 target_link_libraries(
-  Yolo12DetectionDemo PUBLIC ${link_libraries} ${OpenCV_LIBS} executorch_core
+  Yolo26DetectionDemo PUBLIC ${link_libraries} ${OpenCV_LIBS} executorch_core
                              extension_module extension_tensor
 )
 
 find_package(Threads REQUIRED)
-target_link_libraries(Yolo12DetectionDemo PRIVATE Threads::Threads)
+target_link_libraries(Yolo26DetectionDemo PRIVATE Threads::Threads)
 target_include_directories(
-  Yolo12DetectionDemo PUBLIC ${_common_include_directories}
+  Yolo26DetectionDemo PUBLIC ${_common_include_directories}
 )

examples/models/yolo12/CMakeLists.txt examples/models/yolo26/CMakeLists.txtexamples/models/yolo12/CMakeLists.txt renamed to examples/models/yolo26/CMakeLists.txt

@@ -1,22 +1,10 @@
-# YOLO12 Detection C++ Inference with ExecuTorch
+# YOLO26 Detection C++ Inference with ExecuTorch
 
-This example demonstrates how to perform inference of [YOLO12 family](https://docs.ultralytics.com/models/yolo12/) detection models in C++ leveraging the Executorch backends:
+This example demonstrates how to perform inference of [YOLO26 family](https://docs.ultralytics.com/models/yolo26/) detection models in C++ leveraging the Executorch backends:
 
 - [OpenVINO](../../../backends/openvino/README.md)
 - [XNNPACK](../../../backends/xnnpack/README.md)
 
-## Performance Evaluation
-
-| CPU                            | Model   | Backend  | Device | Precision | Average Latency, ms |
-|--------------------------------|---------|----------|--------|-----------|---------------------|
-| Intel(R) Core(TM) Ultra 7 155H | yolo12s | openvino | CPU    | FP32      | 88.3549             |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12s | openvino | CPU    | INT8      | 53.066              |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12l | openvino | CPU    | FP32      | 317.953             |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12l | openvino | CPU    | INT8      | 150.846             |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12s | openvino | GPU    | FP32      | 32.71               |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12l | openvino | GPU    | FP32      | 70.885              |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12s | xnnpack  | CPU    | FP32      | 169.36              |
-| Intel(R) Core(TM) Ultra 7 155H | yolo12l | xnnpack  | CPU    | FP32      | 436.876             |
 
 ## Instructions
 
@@ -38,32 +26,32 @@ The demo requires the `ultralytics` package, which depends on `torch` and `torch
 python -m pip install --upgrade-strategy only-if-needed -r requirements.txt
 ```
 
-### Step 4: Export the YOLO12 model to the ExecuTorch
+### Step 4: Export the YOLO26 model to the ExecuTorch
 
 OpenVINO:
 
 ```bash
-python export_and_validate.py --model_name yolo12s --input_dims=[1920,1080]  --backend openvino --device CPU
+python export_and_validate.py --model_name yolo26s --input_dims=[1920,1080]  --backend openvino --device CPU
 ```
 
 OpenVINO quantized model:
 
 ```bash
-python export_and_validate.py --model_name yolo12s --input_dims=[1920,1080]  --backend openvino --quantize --video_path /path/to/calibration/video --device CPU
+python export_and_validate.py --model_name yolo26s --input_dims=[1920,1080]  --backend openvino --quantize --video_path /path/to/calibration/video --device CPU
 ```
 
 XNNPACK:
 
 ```bash
-python export_and_validate.py --model_name yolo12s --input_dims=[1920,1080] --backend xnnpack
+python export_and_validate.py --model_name yolo26s --input_dims=[1920,1080] --backend xnnpack
 ```
 
 > **_NOTE:_**  Quantization for XNNPACK backend is WIP. Please refere to <https://github.com/pytorch/executorch/issues/11523> for more details.
 
 Exported model could be validated using the `--validate` key:
 
 ```bash
-python export_and_validate.py --model_name yolo12s --backend ... --validate dataset_name.yaml
+python export_and_validate.py --model_name yolo26s --backend ... --validate dataset_name.yaml
 ```
 
 A list of available datasets and instructions on how to use a custom dataset can be found [here](https://docs.ultralytics.com/datasets/detect/).
@@ -80,7 +68,7 @@ python export_and_validate.py --help
 OpenVINO:
 
 ```bash
-cd examples/models/yolo12
+cd examples/models/yolo26
 mkdir build && cd build
 cmake -DCMAKE_BUILD_TYPE=Release -DUSE_OPENVINO_BACKEND=ON ..
 make -j$(nproc)
@@ -89,7 +77,7 @@ make -j$(nproc)
 XNNPACK:
 
 ```bash
-cd examples/models/yolo12
+cd examples/models/yolo26
 mkdir build && cd build
 cmake -DCMAKE_BUILD_TYPE=Release -DUSE_XNNPACK_BACKEND=ON ..
 make -j$(nproc)
@@ -98,13 +86,13 @@ make -j$(nproc)
 ### Step 6: Run the demo
 
 ```bash
-./build/Yolo12DetectionDemo -model_path /path/to/exported/model -input_path /path/to/video/file -output_path /path/to/output/annotated/video
+./build/Yolo26DetectionDemo -model_path /path/to/exported/model -input_path /path/to/video/file -output_path /path/to/output/annotated/video
 ```
 
 To get a full parameters description please use the following command:
 
 ```bash
-./build/Yolo12DetectionDemo --help
+./build/Yolo26DetectionDemo --help
 ```
 
 ## Credits

examples/models/yolo12/README.md examples/models/yolo26/README.mdexamples/models/yolo12/README.md renamed to examples/models/yolo26/README.md

@@ -76,52 +76,50 @@ def lower_to_openvino(
     subset_size: int,
     quantize: bool,
 ) -> ExecutorchProgramManager:
-    # Import openvino locally to avoid nncf side-effects
-    import nncf.torch
     from executorch.backends.openvino.partitioner import OpenvinoPartitioner
     from executorch.backends.openvino.quantizer import OpenVINOQuantizer
     from executorch.backends.openvino.quantizer.quantizer import QuantizationMode
+    import nncf
     from nncf.experimental.torch.fx import quantize_pt2e
 
-    with nncf.torch.disable_patching():
-        if quantize:
-            target_input_dims = tuple(example_args[0].shape[2:])
-
-            def ext_transform_fn(sample):
-                sample = transform_fn(sample)
-                return pad_to_target(sample, target_input_dims)
+    if quantize:
+        target_input_dims = tuple(example_args[0].shape[2:])
 
-            quantizer = OpenVINOQuantizer(mode=QuantizationMode.INT8_TRANSFORMER)
-            quantizer.set_ignored_scope(
-                types=["mul", "sub", "sigmoid", "__getitem__"],
-            )
-            quantized_model = quantize_pt2e(
-                aten_dialect.module(),
-                quantizer,
-                nncf.Dataset(calibration_dataset, ext_transform_fn),
-                subset_size=subset_size,
-                smooth_quant=True,
-                fold_quantize=False,
-            )
+        def ext_transform_fn(sample):
+            sample = transform_fn(sample)
+            return pad_to_target(sample, target_input_dims)
 
-            aten_dialect = torch.export.export(quantized_model, example_args)
-            # Convert to edge dialect and lower the module to the backend with a custom partitioner
-        compile_spec = [CompileSpec("device", device.encode())]
-        lowered_module: EdgeProgramManager = to_edge_transform_and_lower(
-            aten_dialect,
-            partitioner=[
-                OpenvinoPartitioner(compile_spec),
-            ],
-            compile_config=EdgeCompileConfig(
-                _skip_dim_order=True,
-            ),
+        quantizer = OpenVINOQuantizer(mode=QuantizationMode.INT8_MIXED)
+        quantizer.set_ignored_scope(
+            subgraphs=[(["detach", "detach_1", "detach_2"], ["output"])]
         )
-
-        # Apply backend-specific passes
-        return lowered_module.to_executorch(
-            config=executorch.exir.ExecutorchBackendConfig()
+        quantized_model = quantize_pt2e(
+            aten_dialect.module(),
+            quantizer,
+            nncf.Dataset(calibration_dataset, ext_transform_fn),
+            subset_size=subset_size,
+            fast_bias_correction=True,
+            fold_quantize=False,
         )
 
+        aten_dialect = torch.export.export(quantized_model, example_args)
+        # Convert to edge dialect and lower the module to the backend with a custom partitioner
+    compile_spec = [CompileSpec("device", device.encode())]
+    lowered_module: EdgeProgramManager = to_edge_transform_and_lower(
+        aten_dialect,
+        partitioner=[
+            OpenvinoPartitioner(compile_spec),
+        ],
+        compile_config=EdgeCompileConfig(
+            _skip_dim_order=True,
+        ),
+    )
+
+    # Apply backend-specific passes
+    return lowered_module.to_executorch(
+        config=executorch.exir.ExecutorchBackendConfig()
+    )
+
 
 def lower_to_xnnpack(
     aten_dialect: ExportedProgram,
@@ -208,7 +206,7 @@ def main(
     Main function to load, quantize, and export an Yolo model model.
 
     :param model_name: The name of the YOLO model to load.
-    :param input_dims: Input dims to use for the export of a YOLO12 model.
+    :param input_dims: Input dims to use for the export of a YOLO26 model.
     :param quantize: Whether to quantize the model.
     :param video_path: Path to the video to use for the calibration
     :param subset_size: Subset size for the quantized model calibration. The default value is 300.
@@ -272,7 +270,7 @@ def transform_fn(frame):
         if input_dims != [640, 640]:
             raise NotImplementedError(
                 f"Validation with the custom input shape {input_dims} is not implmenented."
-                " Please use the default --input_dims=[640, 640] for the validation."
+                " Please use the default --input_dims=[640,640] for the validation."
             )
         stats = validate_yolo(model, exec_prog, val_dataset_yaml_path)
         for stat, value in stats.items():
@@ -290,10 +288,13 @@ def _prepare_validation(
     }  # highest priority args on the right
 
     validator = model._smart_load("validator")(args=args, _callbacks=model.callbacks)
+    validator.device = torch.device("cpu")
     stride = 32  # default stride
     validator.stride = stride  # used in get_dataloader() for padding
     validator.data = check_det_dataset(dataset_yaml_path)
     validator.init_metrics(unwrap_model(model))
+    validator.device = torch.device("cpu")
+    validator.end2end = False
 
     data_loader = validator.get_dataloader(
         validator.data.get(validator.args.split), validator.args.batch
@@ -337,9 +338,9 @@ def validate_yolo(
     parser.add_argument(
         "--model_name",
         type=str,
-        default="yolo12s",
-        choices=["yolo12n", "yolo12s", "yolo12m", "yolo12l", "yolo12x"],
-        help="Ultralytics yolo12 model name.",
+        default="yolo26s",
+        choices=["yolo26n", "yolo26s", "yolo26m", "yolo26l", "yolo26x"],
+        help="Ultralytics yolo26 model name.",
     )
     parser.add_argument(
         "--input_dims",