Refactor with simplifications (#19329)

Co-authored-by: UltralyticsAssistant <web@ultralytics.com>

examples/YOLOv8-Segmentation-ONNXRuntime-Python/main.py

@@ -79,9 +79,7 @@ class YOLOv8Seg:
         predictions = self.session.run(None, {self.session.get_inputs()[0].name: processed_image})
 
         # Post-process
-        results = self.postprocess(im0, processed_image, predictions)
+        return self.postprocess(im0, processed_image, predictions)
-
-        return results
 
     def preprocess(self, image, new_shape: Union[Tuple, List] = (640, 640)):
         """
@@ -99,8 +97,7 @@ class YOLOv8Seg:
         """
         image, _, _ = self.__resize_and_pad_image(image=image, new_shape=new_shape)
         image = self.__reshape_image(image=image)
-        processed_image = image[None] if len(image.shape) == 3 else image
+        return image[None] if len(image.shape) == 3 else image
-        return processed_image
 
     def __reshape_image(self, image: np.ndarray) -> np.ndarray:
         """
@@ -117,8 +114,7 @@ class YOLOv8Seg:
         """
         image = image.transpose([2, 0, 1])
         image = image[np.newaxis, ...]
-        image = np.ascontiguousarray(image).astype(np.float32) / 255
+        return np.ascontiguousarray(image).astype(np.float32) / 255
-        return image
 
     def __resize_and_pad_image(
         self, image=np.ndarray, new_shape: Union[Tuple, List] = (640, 640), color: Union[Tuple, List] = (114, 114, 114)

ultralytics/engine/exporter.py

@@ -309,9 +309,8 @@ class Exporter:
                 "WARNING ⚠️ INT8 export requires a missing 'data' arg for calibration. "
                 f"Using default 'data={self.args.data}'."
             )
-        if tfjs:
+        if tfjs and (ARM64 and LINUX):
-            if ARM64 and LINUX:
+            raise SystemError("TensorFlow.js export not supported on ARM64 Linux")
-                raise SystemError("TensorFlow.js export not supported on ARM64 Linux")
 
         # Input
         im = torch.zeros(self.args.batch, 3, *self.imgsz).to(self.device)

ultralytics/nn/autobackend.py

@@ -197,12 +197,13 @@ class AutoBackend(nn.Module):
             import onnxruntime
 
             providers = ["CPUExecutionProvider"]
-            if cuda and "CUDAExecutionProvider" in onnxruntime.get_available_providers():
+            if cuda:
-                providers.insert(0, "CUDAExecutionProvider")
+                if "CUDAExecutionProvider" in onnxruntime.get_available_providers():
-            elif cuda:  # Only log warning if CUDA was requested but unavailable
+                    providers.insert(0, "CUDAExecutionProvider")
-                LOGGER.warning("WARNING ⚠️ Failed to start ONNX Runtime with CUDA. Using CPU...")
+                else:  # Only log warning if CUDA was requested but unavailable
-                device = torch.device("cpu")
+                    LOGGER.warning("WARNING ⚠️ Failed to start ONNX Runtime with CUDA. Using CPU...")
-                cuda = False
+                    device = torch.device("cpu")
+                    cuda = False
             LOGGER.info(f"Using ONNX Runtime {providers[0]}")
             if onnx:
                 session = onnxruntime.InferenceSession(w, providers=providers)
@@ -223,7 +224,7 @@ class AutoBackend(nn.Module):
             output_names = [x.name for x in session.get_outputs()]
             metadata = session.get_modelmeta().custom_metadata_map
             dynamic = isinstance(session.get_outputs()[0].shape[0], str)
-            fp16 = True if "float16" in session.get_inputs()[0].type else False
+            fp16 = "float16" in session.get_inputs()[0].type
             if not dynamic:
                 io = session.io_binding()
                 bindings = []

ultralytics/utils/torch_utils.py

@@ -317,8 +317,7 @@ def model_info(model, detailed=False, verbose=True, imgsz=640):
             if len(m._parameters):
                 for pn, p in m.named_parameters():
                     LOGGER.info(
-                        f"{i:>5g}{mn + '.' + pn:>40}{mt:>20}{p.requires_grad!r:>10}{p.numel():>12g}"
+                        f"{i:>5g}{f'{mn}.{pn}':>40}{mt:>20}{p.requires_grad!r:>10}{p.numel():>12g}{str(list(p.shape)):>20}{p.mean():>10.3g}{p.std():>10.3g}{str(p.dtype).replace('torch.', ''):>15}"
-                        f"{str(list(p.shape)):>20}{p.mean():>10.3g}{p.std():>10.3g}{str(p.dtype).replace('torch.', ''):>15}"
                     )
             else:  # layers with no learnable params
                 LOGGER.info(f"{i:>5g}{mn:>40}{mt:>20}{False!r:>10}{0:>12g}{str([]):>20}{'-':>10}{'-':>10}{'-':>15}")