dgenerate-ultralytics-headless 8.3.194__py3-none-any.whl → 8.3.196__py3-none-any.whl

ultralytics/engine/results.py

@@ -30,7 +30,7 @@ class BaseTensor(SimpleClass):
 
     Attributes:
         data (torch.Tensor | np.ndarray): Prediction data such as bounding boxes, masks, or keypoints.
-        orig_shape (Tuple[int, int]): Original shape of the image, typically in the format (height, width).
+        orig_shape (tuple[int, int]): Original shape of the image, typically in the format (height, width).
 
     Methods:
         cpu: Return a copy of the tensor stored in CPU memory.
@@ -54,7 +54,7 @@ class BaseTensor(SimpleClass):
 
         Args:
             data (torch.Tensor | np.ndarray): Prediction data such as bounding boxes, masks, or keypoints.
-            orig_shape (Tuple[int, int]): Original shape of the image in (height, width) format.
+            orig_shape (tuple[int, int]): Original shape of the image in (height, width) format.
 
         Examples:
             >>> import torch
@@ -72,7 +72,7 @@ class BaseTensor(SimpleClass):
         Return the shape of the underlying data tensor.
 
         Returns:
-            (Tuple[int, ...]): The shape of the data tensor.
+            (tuple[int, ...]): The shape of the data tensor.
 
         Examples:
             >>> data = torch.rand(100, 4)
@@ -174,7 +174,7 @@ class BaseTensor(SimpleClass):
         Return a new BaseTensor instance containing the specified indexed elements of the data tensor.
 
         Args:
-            idx (int | List[int] | torch.Tensor): Index or indices to select from the data tensor.
+            idx (int | list[int] | torch.Tensor): Index or indices to select from the data tensor.
 
         Returns:
             (BaseTensor): A new BaseTensor instance containing the indexed data.
@@ -199,7 +199,7 @@ class Results(SimpleClass, DataExportMixin):
 
     Attributes:
         orig_img (np.ndarray): The original image as a numpy array.
-        orig_shape (Tuple[int, int]): Original image shape in (height, width) format.
+        orig_shape (tuple[int, int]): Original image shape in (height, width) format.
         boxes (Boxes | None): Detected bounding boxes.
         masks (Masks | None): Segmentation masks.
         probs (Probs | None): Classification probabilities.
@@ -261,7 +261,7 @@ class Results(SimpleClass, DataExportMixin):
             probs (torch.Tensor | None): A 1D tensor of probabilities of each class for classification task.
             keypoints (torch.Tensor | None): A 2D tensor of keypoint coordinates for each detection.
             obb (torch.Tensor | None): A 2D tensor of oriented bounding box coordinates for each detection.
-            speed (Dict | None): A dictionary containing preprocess, inference, and postprocess speeds (ms/image).
+            speed (dict | None): A dictionary containing preprocess, inference, and postprocess speeds (ms/image).
 
         Examples:
             >>> results = model("path/to/image.jpg")
@@ -799,7 +799,7 @@ class Results(SimpleClass, DataExportMixin):
             decimals (int): Number of decimal places to round the output values to.
 
         Returns:
-            (List[Dict[str, Any]]): A list of dictionaries, each containing summarized information for a single detection
+            (list[dict[str, Any]]): A list of dictionaries, each containing summarized information for a single detection
                 or classification result. The structure of each dictionary varies based on the task type
                 (classification or detection) and available information (boxes, masks, keypoints).
 
@@ -862,7 +862,7 @@ class Boxes(BaseTensor):
 
     Attributes:
         data (torch.Tensor | np.ndarray): The raw tensor containing detection boxes and associated data.
-        orig_shape (Tuple[int, int]): The original image dimensions (height, width).
+        orig_shape (tuple[int, int]): The original image dimensions (height, width).
         is_track (bool): Indicates whether tracking IDs are included in the box data.
         xyxy (torch.Tensor | np.ndarray): Boxes in [x1, y1, x2, y2] format.
         conf (torch.Tensor | np.ndarray): Confidence scores for each box.
@@ -901,11 +901,11 @@ class Boxes(BaseTensor):
             boxes (torch.Tensor | np.ndarray): A tensor or numpy array with detection boxes of shape
                 (num_boxes, 6) or (num_boxes, 7). Columns should contain
                 [x1, y1, x2, y2, confidence, class, (optional) track_id].
-            orig_shape (Tuple[int, int]): The original image shape as (height, width). Used for normalization.
+            orig_shape (tuple[int, int]): The original image shape as (height, width). Used for normalization.
 
         Attributes:
             data (torch.Tensor): The raw tensor containing detection boxes and their associated data.
-            orig_shape (Tuple[int, int]): The original image size, used for normalization.
+            orig_shape (tuple[int, int]): The original image size, used for normalization.
             is_track (bool): Indicates whether tracking IDs are included in the box data.
 
         Examples:
@@ -1081,8 +1081,8 @@ class Masks(BaseTensor):
     Attributes:
         data (torch.Tensor | np.ndarray): The raw tensor or array containing mask data.
         orig_shape (tuple): Original image shape in (height, width) format.
-        xy (List[np.ndarray]): A list of segments in pixel coordinates.
-        xyn (List[np.ndarray]): A list of normalized segments.
+        xy (list[np.ndarray]): A list of segments in pixel coordinates.
+        xyn (list[np.ndarray]): A list of normalized segments.
 
     Methods:
         cpu: Return a copy of the Masks object with the mask tensor on CPU memory.
@@ -1127,7 +1127,7 @@ class Masks(BaseTensor):
         are normalized relative to the original image shape.
 
         Returns:
-            (List[np.ndarray]): A list of numpy arrays, where each array contains the normalized xy-coordinates
+            (list[np.ndarray]): A list of numpy arrays, where each array contains the normalized xy-coordinates
                 of a single segmentation mask. Each array has shape (N, 2), where N is the number of points in the
                 mask contour.
 
@@ -1152,7 +1152,7 @@ class Masks(BaseTensor):
         Masks object. The coordinates are scaled to match the original image dimensions.
 
         Returns:
-            (List[np.ndarray]): A list of numpy arrays, where each array contains the [x, y] pixel
+            (list[np.ndarray]): A list of numpy arrays, where each array contains the [x, y] pixel
                 coordinates for a single segmentation mask. Each array has shape (N, 2), where N is the
                 number of points in the segment.
 
@@ -1179,7 +1179,7 @@ class Keypoints(BaseTensor):
 
     Attributes:
         data (torch.Tensor): The raw tensor containing keypoint data.
-        orig_shape (Tuple[int, int]): The original image dimensions (height, width).
+        orig_shape (tuple[int, int]): The original image dimensions (height, width).
         has_visible (bool): Indicates whether visibility information is available for keypoints.
         xy (torch.Tensor): Keypoint coordinates in [x, y] format.
         xyn (torch.Tensor): Normalized keypoint coordinates in [x, y] format, relative to orig_shape.
@@ -1213,7 +1213,7 @@ class Keypoints(BaseTensor):
             keypoints (torch.Tensor): A tensor containing keypoint data. Shape can be either:
                 - (num_objects, num_keypoints, 2) for x, y coordinates only
                 - (num_objects, num_keypoints, 3) for x, y coordinates and confidence scores
-            orig_shape (Tuple[int, int]): The original image dimensions (height, width).
+            orig_shape (tuple[int, int]): The original image dimensions (height, width).
 
         Examples:
             >>> kpts = torch.rand(1, 17, 3)  # 1 object, 17 keypoints (COCO format), x,y,conf
@@ -1301,7 +1301,7 @@ class Probs(BaseTensor):
         data (torch.Tensor | np.ndarray): The raw tensor or array containing classification probabilities.
         orig_shape (tuple | None): The original image shape as (height, width). Not used in this class.
         top1 (int): Index of the class with the highest probability.
-        top5 (List[int]): Indices of the top 5 classes by probability.
+        top5 (list[int]): Indices of the top 5 classes by probability.
         top1conf (torch.Tensor | np.ndarray): Confidence score of the top 1 class.
         top5conf (torch.Tensor | np.ndarray): Confidence scores of the top 5 classes.
 
@@ -1339,7 +1339,7 @@ class Probs(BaseTensor):
         Attributes:
             data (torch.Tensor | np.ndarray): The raw tensor or array containing classification probabilities.
             top1 (int): Index of the top 1 class.
-            top5 (List[int]): Indices of the top 5 classes.
+            top5 (list[int]): Indices of the top 5 classes.
             top1conf (torch.Tensor | np.ndarray): Confidence of the top 1 class.
             top5conf (torch.Tensor | np.ndarray): Confidences of the top 5 classes.
 
@@ -1379,7 +1379,7 @@ class Probs(BaseTensor):
         Return the indices of the top 5 class probabilities.
 
         Returns:
-            (List[int]): A list containing the indices of the top 5 class probabilities, sorted in descending order.
+            (list[int]): A list containing the indices of the top 5 class probabilities, sorted in descending order.
 
         Examples:
             >>> probs = Probs(torch.tensor([0.1, 0.2, 0.3, 0.4, 0.5]))
@@ -1476,11 +1476,11 @@ class OBB(BaseTensor):
             boxes (torch.Tensor | np.ndarray): A tensor or numpy array containing the detection boxes,
                 with shape (num_boxes, 7) or (num_boxes, 8). The last two columns contain confidence and class values.
                 If present, the third last column contains track IDs, and the fifth column contains rotation.
-            orig_shape (Tuple[int, int]): Original image size, in the format (height, width).
+            orig_shape (tuple[int, int]): Original image size, in the format (height, width).
 
         Attributes:
             data (torch.Tensor | np.ndarray): The raw OBB tensor.
-            orig_shape (Tuple[int, int]): The original image shape.
+            orig_shape (tuple[int, int]): The original image shape.
             is_track (bool): Whether the boxes include tracking IDs.
 
         Raises:

ultralytics/engine/trainer.py

@@ -46,6 +46,7 @@ from ultralytics.utils.torch_utils import (
     TORCH_2_4,
     EarlyStopping,
     ModelEMA,
+    attempt_compile,
     autocast,
     convert_optimizer_state_dict_to_fp16,
     init_seeds,
@@ -54,6 +55,7 @@ from ultralytics.utils.torch_utils import (
     strip_optimizer,
     torch_distributed_zero_first,
     unset_deterministic,
+    unwrap_model,
 )
 
 
@@ -256,6 +258,14 @@ class BaseTrainer:
         self.model = self.model.to(self.device)
         self.set_model_attributes()
 
+        # Initialize loss criterion before compilation for torch.compile compatibility
+        if hasattr(self.model, "init_criterion"):
+            self.model.criterion = self.model.init_criterion()
+
+        # Compile model
+        if self.args.compile:
+            self.model = attempt_compile(self.model, device=self.device)
+
         # Freeze layers
         freeze_list = (
             self.args.freeze
@@ -404,6 +414,7 @@ class BaseTrainer:
                 # Forward
                 with autocast(self.amp):
                     batch = self.preprocess_batch(batch)
+                    metadata = {k: batch.pop(k, None) for k in ["im_file", "ori_shape", "resized_shape"]}
                     loss, self.loss_items = self.model(batch)
                     self.loss = loss.sum()
                     if RANK != -1:
@@ -445,6 +456,7 @@ class BaseTrainer:
                     )
                     self.run_callbacks("on_batch_end")
                     if self.args.plots and ni in self.plot_idx:
+                        batch = {**batch, **metadata}
                         self.plot_training_samples(batch, ni)
 
                 self.run_callbacks("on_train_batch_end")
@@ -565,7 +577,7 @@ class BaseTrainer:
                 "epoch": self.epoch,
                 "best_fitness": self.best_fitness,
                 "model": None,  # resume and final checkpoints derive from EMA
-                "ema": deepcopy(self.ema.ema).half(),
+                "ema": deepcopy(unwrap_model(self.ema.ema)).half(),
                 "updates": self.ema.updates,
                 "optimizer": convert_optimizer_state_dict_to_fp16(deepcopy(self.optimizer.state_dict())),
                 "train_args": vars(self.args),  # save as dict
@@ -592,8 +604,6 @@ class BaseTrainer:
             self.best.write_bytes(serialized_ckpt)  # save best.pt
         if (self.save_period > 0) and (self.epoch % self.save_period == 0):
             (self.wdir / f"epoch{self.epoch}.pt").write_bytes(serialized_ckpt)  # save epoch, i.e. 'epoch3.pt'
-        # if self.args.close_mosaic and self.epoch == (self.epochs - self.args.close_mosaic - 1):
-        #    (self.wdir / "last_mosaic.pt").write_bytes(serialized_ckpt)  # save mosaic checkpoint
 
     def get_dataset(self):
         """
@@ -667,7 +677,7 @@ class BaseTrainer:
 
     def validate(self):
         """
-        Run validation on test set using self.validator.
+        Run validation on val set using self.validator.
 
         Returns:
             metrics (dict): Dictionary of validation metrics.
@@ -755,6 +765,7 @@ class BaseTrainer:
                     strip_optimizer(f, updates={k: ckpt[k]} if k in ckpt else None)
                     LOGGER.info(f"\nValidating {f}...")
                     self.validator.args.plots = self.args.plots
+                    self.validator.args.compile = False  # disable final val compile as too slow
                     self.metrics = self.validator(model=f)
                     self.metrics.pop("fitness", None)
                     self.run_callbacks("on_fit_epoch_end")

ultralytics/engine/validator.py

@@ -36,7 +36,7 @@ from ultralytics.nn.autobackend import AutoBackend
 from ultralytics.utils import LOGGER, TQDM, callbacks, colorstr, emojis
 from ultralytics.utils.checks import check_imgsz
 from ultralytics.utils.ops import Profile
-from ultralytics.utils.torch_utils import de_parallel, select_device, smart_inference_mode
+from ultralytics.utils.torch_utils import attempt_compile, select_device, smart_inference_mode, unwrap_model
 
 
 class BaseValidator:
@@ -148,6 +148,8 @@ class BaseValidator:
             # Force FP16 val during training
             self.args.half = self.device.type != "cpu" and trainer.amp
             model = trainer.ema.ema or trainer.model
+            if trainer.args.compile and hasattr(model, "_orig_mod"):
+                model = model._orig_mod  # validate non-compiled original model to avoid issues
             model = model.half() if self.args.half else model.float()
             self.loss = torch.zeros_like(trainer.loss_items, device=trainer.device)
             self.args.plots &= trainer.stopper.possible_stop or (trainer.epoch == trainer.epochs - 1)
@@ -186,6 +188,8 @@ class BaseValidator:
             self.dataloader = self.dataloader or self.get_dataloader(self.data.get(self.args.split), self.args.batch)
 
             model.eval()
+            if self.args.compile:
+                model = attempt_compile(model, device=self.device)
             model.warmup(imgsz=(1 if pt else self.args.batch, self.data["channels"], imgsz, imgsz))  # warmup
 
         self.run_callbacks("on_val_start")
@@ -196,7 +200,7 @@ class BaseValidator:
             Profile(device=self.device),
         )
         bar = TQDM(self.dataloader, desc=self.get_desc(), total=len(self.dataloader))
-        self.init_metrics(de_parallel(model))
+        self.init_metrics(unwrap_model(model))
         self.jdict = []  # empty before each val
         for batch_i, batch in enumerate(bar):
             self.run_callbacks("on_val_batch_start")

ultralytics/hub/google/__init__.py

@@ -15,7 +15,7 @@ class GCPRegions:
     geographical location, tier classification, and network latency.
 
     Attributes:
-        regions (Dict[str, Tuple[int, str, str]]): A dictionary of GCP regions with their tier, city, and country.
+        regions (dict[str, tuple[int, str, str]]): A dictionary of GCP regions with their tier, city, and country.
 
     Methods:
         tier1: Returns a list of tier 1 GCP regions.
@@ -136,7 +136,7 @@ class GCPRegions:
             attempts (int, optional): Number of ping attempts per region.
 
         Returns:
-            (List[Tuple[str, float, float, float, float]]): List of tuples containing region information and
+            (list[tuple[str, float, float, float, float]]): List of tuples containing region information and
                 latency statistics. Each tuple contains (region, mean_latency, std_dev, min_latency, max_latency).
 
         Examples:

ultralytics/hub/session.py

@@ -28,13 +28,13 @@ class HUBTrainingSession:
     Attributes:
         model_id (str): Identifier for the YOLO model being trained.
         model_url (str): URL for the model in Ultralytics HUB.
-        rate_limits (Dict[str, int]): Rate limits for different API calls in seconds.
-        timers (Dict[str, Any]): Timers for rate limiting.
-        metrics_queue (Dict[str, Any]): Queue for the model's metrics.
-        metrics_upload_failed_queue (Dict[str, Any]): Queue for metrics that failed to upload.
+        rate_limits (dict[str, int]): Rate limits for different API calls in seconds.
+        timers (dict[str, Any]): Timers for rate limiting.
+        metrics_queue (dict[str, Any]): Queue for the model's metrics.
+        metrics_upload_failed_queue (dict[str, Any]): Queue for metrics that failed to upload.
         model (Any): Model data fetched from Ultralytics HUB.
         model_file (str): Path to the model file.
-        train_args (Dict[str, Any]): Arguments for training the model.
+        train_args (dict[str, Any]): Arguments for training the model.
         client (Any): Client for interacting with Ultralytics HUB.
         filename (str): Filename of the model.
 
@@ -98,7 +98,7 @@ class HUBTrainingSession:
 
         Args:
             identifier (str): Model identifier used to initialize the HUB training session.
-            args (Dict[str, Any], optional): Arguments for creating a new model if identifier is not a HUB model URL.
+            args (dict[str, Any], optional): Arguments for creating a new model if identifier is not a HUB model URL.
 
         Returns:
             session (HUBTrainingSession | None): An authenticated session or None if creation fails.
@@ -144,7 +144,7 @@ class HUBTrainingSession:
         Initialize a HUB training session with the specified model arguments.
 
         Args:
-            model_args (Dict[str, Any]): Arguments for creating the model, including batch size, epochs, image size,
+            model_args (dict[str, Any]): Arguments for creating the model, including batch size, epochs, image size,
                 etc.
 
         Returns:

ultralytics/models/fastsam/model.py

@@ -63,14 +63,14 @@ class FastSAM(Model):
             source (str | PIL.Image | np.ndarray): Input source for prediction, can be a file path, URL, PIL image,
                 or numpy array.
             stream (bool): Whether to enable real-time streaming mode for video inputs.
-            bboxes (List, optional): Bounding box coordinates for prompted segmentation in format [[x1, y1, x2, y2]].
-            points (List, optional): Point coordinates for prompted segmentation in format [[x, y]].
-            labels (List, optional): Class labels for prompted segmentation.
-            texts (List, optional): Text prompts for segmentation guidance.
+            bboxes (list, optional): Bounding box coordinates for prompted segmentation in format [[x1, y1, x2, y2]].
+            points (list, optional): Point coordinates for prompted segmentation in format [[x, y]].
+            labels (list, optional): Class labels for prompted segmentation.
+            texts (list, optional): Text prompts for segmentation guidance.
             **kwargs (Any): Additional keyword arguments passed to the predictor.
 
         Returns:
-            (List): List of Results objects containing the prediction results.
+            (list): List of Results objects containing the prediction results.
         """
         prompts = dict(bboxes=bboxes, points=points, labels=labels, texts=texts)
         return super().predict(source, stream, prompts=prompts, **kwargs)

ultralytics/models/fastsam/predict.py

@@ -52,12 +52,12 @@ class FastSAMPredictor(SegmentationPredictor):
         Apply postprocessing to FastSAM predictions and handle prompts.
 
         Args:
-            preds (List[torch.Tensor]): Raw predictions from the model.
+            preds (list[torch.Tensor]): Raw predictions from the model.
             img (torch.Tensor): Input image tensor that was fed to the model.
-            orig_imgs (List[np.ndarray]): Original images before preprocessing.
+            orig_imgs (list[np.ndarray]): Original images before preprocessing.
 
         Returns:
-            (List[Results]): Processed results with prompts applied.
+            (list[Results]): Processed results with prompts applied.
         """
         bboxes = self.prompts.pop("bboxes", None)
         points = self.prompts.pop("points", None)
@@ -80,14 +80,14 @@ class FastSAMPredictor(SegmentationPredictor):
         Perform image segmentation inference based on cues like bounding boxes, points, and text prompts.
 
         Args:
-            results (Results | List[Results]): Original inference results from FastSAM models without any prompts.
-            bboxes (np.ndarray | List, optional): Bounding boxes with shape (N, 4), in XYXY format.
-            points (np.ndarray | List, optional): Points indicating object locations with shape (N, 2), in pixels.
-            labels (np.ndarray | List, optional): Labels for point prompts, shape (N, ). 1 = foreground, 0 = background.
-            texts (str | List[str], optional): Textual prompts, a list containing string objects.
+            results (Results | list[Results]): Original inference results from FastSAM models without any prompts.
+            bboxes (np.ndarray | list, optional): Bounding boxes with shape (N, 4), in XYXY format.
+            points (np.ndarray | list, optional): Points indicating object locations with shape (N, 2), in pixels.
+            labels (np.ndarray | list, optional): Labels for point prompts, shape (N, ). 1 = foreground, 0 = background.
+            texts (str | list[str], optional): Textual prompts, a list containing string objects.
 
         Returns:
-            (List[Results]): Output results filtered and determined by the provided prompts.
+            (list[Results]): Output results filtered and determined by the provided prompts.
         """
         if bboxes is None and points is None and texts is None:
             return results
@@ -154,8 +154,8 @@ class FastSAMPredictor(SegmentationPredictor):
         Perform CLIP inference to calculate similarity between images and text prompts.
 
         Args:
-            images (List[PIL.Image]): List of source images, each should be PIL.Image with RGB channel order.
-            texts (List[str]): List of prompt texts, each should be a string object.
+            images (list[PIL.Image]): List of source images, each should be PIL.Image with RGB channel order.
+            texts (list[str]): List of prompt texts, each should be a string object.
 
         Returns:
             (torch.Tensor): Similarity matrix between given images and texts with shape (M, N).

ultralytics/models/nas/model.py

@@ -91,7 +91,7 @@ class NAS(Model):
             verbose (bool): Controls verbosity.
 
         Returns:
-            (Dict[str, Any]): Model information dictionary.
+            (dict[str, Any]): Model information dictionary.
         """
         return model_info(self.model, detailed=detailed, verbose=verbose, imgsz=640)
 

ultralytics/models/rtdetr/predict.py

@@ -47,7 +47,7 @@ class RTDETRPredictor(BasePredictor):
             orig_imgs (list | torch.Tensor): Original, unprocessed images.
 
         Returns:
-            results (List[Results]): A list of Results objects containing the post-processed bounding boxes,
+            results (list[Results]): A list of Results objects containing the post-processed bounding boxes,
                 confidence scores, and class labels.
         """
         if not isinstance(preds, (list, tuple)):  # list for PyTorch inference but list[0] Tensor for export inference
@@ -82,7 +82,7 @@ class RTDETRPredictor(BasePredictor):
         (640) and scale_filled.
 
         Args:
-            im (List[np.ndarray]  | torch.Tensor): Input images of shape (N, 3, H, W) for tensor,
+            im (list[np.ndarray]  | torch.Tensor): Input images of shape (N, 3, H, W) for tensor,
                 [(H, W, 3) x N] for list.
 
         Returns:

ultralytics/models/rtdetr/val.py

@@ -163,11 +163,11 @@ class RTDETRValidator(DetectionValidator):
         Apply Non-maximum suppression to prediction outputs.
 
         Args:
-            preds (torch.Tensor | List | Tuple): Raw predictions from the model. If tensor, should have shape
+            preds (torch.Tensor | list | tuple): Raw predictions from the model. If tensor, should have shape
                 (batch_size, num_predictions, num_classes + 4) where last dimension contains bbox coords and class scores.
 
         Returns:
-            (List[Dict[str, torch.Tensor]]): List of dictionaries for each image, each containing:
+            (list[dict[str, torch.Tensor]]): List of dictionaries for each image, each containing:
                 - 'bboxes': Tensor of shape (N, 4) with bounding box coordinates
                 - 'conf': Tensor of shape (N,) with confidence scores
                 - 'cls': Tensor of shape (N,) with class indices
@@ -194,9 +194,9 @@ class RTDETRValidator(DetectionValidator):
         Serialize YOLO predictions to COCO json format.
 
         Args:
-            predn (Dict[str, torch.Tensor]): Predictions dictionary containing 'bboxes', 'conf', and 'cls' keys
+            predn (dict[str, torch.Tensor]): Predictions dictionary containing 'bboxes', 'conf', and 'cls' keys
                 with bounding box coordinates, confidence scores, and class predictions.
-            pbatch (Dict[str, Any]): Batch dictionary containing 'imgsz', 'ori_shape', 'ratio_pad', and 'im_file'.
+            pbatch (dict[str, Any]): Batch dictionary containing 'imgsz', 'ori_shape', 'ratio_pad', and 'im_file'.
         """
         path = Path(pbatch["im_file"])
         stem = path.stem

ultralytics/models/sam/amg.py

@@ -19,8 +19,8 @@ def is_box_near_crop_edge(
 
     Args:
         boxes (torch.Tensor): Bounding boxes in XYXY format.
-        crop_box (List[int]): Crop box coordinates in [x0, y0, x1, y1] format.
-        orig_box (List[int]): Original image box coordinates in [x0, y0, x1, y1] format.
+        crop_box (list[int]): Crop box coordinates in [x0, y0, x1, y1] format.
+        orig_box (list[int]): Original image box coordinates in [x0, y0, x1, y1] format.
         atol (float, optional): Absolute tolerance for edge proximity detection.
 
     Returns:
@@ -53,7 +53,7 @@ def batch_iterator(batch_size: int, *args) -> Generator[list[Any]]:
         *args (Any): Variable length input iterables to batch. All iterables must have the same length.
 
     Yields:
-        (List[Any]): A list of batched elements from each input iterable.
+        (list[Any]): A list of batched elements from each input iterable.
 
     Examples:
         >>> data = [1, 2, 3, 4, 5]
@@ -121,13 +121,13 @@ def generate_crop_boxes(
     Generate crop boxes of varying sizes for multiscale image processing, with layered overlapping regions.
 
     Args:
-        im_size (Tuple[int, ...]): Height and width of the input image.
+        im_size (tuple[int, ...]): Height and width of the input image.
         n_layers (int): Number of layers to generate crop boxes for.
         overlap_ratio (float): Ratio of overlap between adjacent crop boxes.
 
     Returns:
-        crop_boxes (List[List[int]]): List of crop boxes in [x0, y0, x1, y1] format.
-        layer_idxs (List[int]): List of layer indices corresponding to each crop box.
+        crop_boxes (list[list[int]]): List of crop boxes in [x0, y0, x1, y1] format.
+        layer_idxs (list[int]): List of layer indices corresponding to each crop box.
 
     Examples:
         >>> im_size = (800, 1200)  # Height, width

ultralytics/models/sam/build.py

@@ -130,10 +130,10 @@ def _build_sam(
     Build a Segment Anything Model (SAM) with specified encoder parameters.
 
     Args:
-        encoder_embed_dim (int | List[int]): Embedding dimension for the encoder.
-        encoder_depth (int | List[int]): Depth of the encoder.
-        encoder_num_heads (int | List[int]): Number of attention heads in the encoder.
-        encoder_global_attn_indexes (List[int] | None): Indexes for global attention in the encoder.
+        encoder_embed_dim (int | list[int]): Embedding dimension for the encoder.
+        encoder_depth (int | list[int]): Depth of the encoder.
+        encoder_num_heads (int | list[int]): Number of attention heads in the encoder.
+        encoder_global_attn_indexes (list[int] | None): Indexes for global attention in the encoder.
         checkpoint (str | None, optional): Path to the model checkpoint file.
         mobile_sam (bool, optional): Whether to build a Mobile-SAM model.
 
@@ -228,12 +228,12 @@ def _build_sam2(
 
     Args:
         encoder_embed_dim (int, optional): Embedding dimension for the encoder.
-        encoder_stages (List[int], optional): Number of blocks in each stage of the encoder.
+        encoder_stages (list[int], optional): Number of blocks in each stage of the encoder.
         encoder_num_heads (int, optional): Number of attention heads in the encoder.
-        encoder_global_att_blocks (List[int], optional): Indices of global attention blocks in the encoder.
-        encoder_backbone_channel_list (List[int], optional): Channel dimensions for each level of the encoder backbone.
-        encoder_window_spatial_size (List[int], optional): Spatial size of the window for position embeddings.
-        encoder_window_spec (List[int], optional): Window specifications for each stage of the encoder.
+        encoder_global_att_blocks (list[int], optional): Indices of global attention blocks in the encoder.
+        encoder_backbone_channel_list (list[int], optional): Channel dimensions for each level of the encoder backbone.
+        encoder_window_spatial_size (list[int], optional): Spatial size of the window for position embeddings.
+        encoder_window_spec (list[int], optional): Window specifications for each stage of the encoder.
         checkpoint (str | None, optional): Path to the checkpoint file for loading pre-trained weights.
 
     Returns:

ultralytics/models/sam/model.py

@@ -91,9 +91,9 @@ class SAM(Model):
             source (str | PIL.Image | np.ndarray): Path to the image or video file, or a PIL.Image object, or
                 a np.ndarray object.
             stream (bool): If True, enables real-time streaming.
-            bboxes (List[List[float]] | None): List of bounding box coordinates for prompted segmentation.
-            points (List[List[float]] | None): List of points for prompted segmentation.
-            labels (List[int] | None): List of labels for prompted segmentation.
+            bboxes (list[list[float]] | None): List of bounding box coordinates for prompted segmentation.
+            points (list[list[float]] | None): List of points for prompted segmentation.
+            labels (list[int] | None): List of labels for prompted segmentation.
             **kwargs (Any): Additional keyword arguments for prediction.
 
         Returns:
@@ -121,9 +121,9 @@ class SAM(Model):
             source (str | PIL.Image | np.ndarray | None): Path to the image or video file, or a PIL.Image
                 object, or a np.ndarray object.
             stream (bool): If True, enables real-time streaming.
-            bboxes (List[List[float]] | None): List of bounding box coordinates for prompted segmentation.
-            points (List[List[float]] | None): List of points for prompted segmentation.
-            labels (List[int] | None): List of labels for prompted segmentation.
+            bboxes (list[list[float]] | None): List of bounding box coordinates for prompted segmentation.
+            points (list[list[float]] | None): List of points for prompted segmentation.
+            labels (list[int] | None): List of labels for prompted segmentation.
             **kwargs (Any): Additional keyword arguments to be passed to the predict method.
 
         Returns:
@@ -160,7 +160,7 @@ class SAM(Model):
         Provide a mapping from the 'segment' task to its corresponding 'Predictor'.
 
         Returns:
-            (Dict[str, Dict[str, Type[Predictor]]]): A dictionary mapping the 'segment' task to its corresponding
+            (dict[str, dict[str, Type[Predictor]]]): A dictionary mapping the 'segment' task to its corresponding
                 Predictor class. For SAM2 models, it maps to SAM2Predictor, otherwise to the standard Predictor.
 
         Examples:

ultralytics/models/sam/modules/blocks.py

@@ -593,7 +593,7 @@ class MultiScaleBlock(nn.Module):
         norm1 (nn.Module): First normalization layer.
         window_size (int): Size of the window for partitioning.
         pool (nn.Module | None): Pooling layer for query downsampling.
-        q_stride (Tuple[int, int] | None): Stride for query pooling.
+        q_stride (tuple[int, int] | None): Stride for query pooling.
         attn (MultiScaleAttention): Multi-scale attention module.
         drop_path (nn.Module): Drop path layer for regularization.
         norm2 (nn.Module): Second normalization layer.
@@ -934,7 +934,7 @@ class Block(nn.Module):
             use_rel_pos (bool): If True, uses relative positional embeddings in attention.
             rel_pos_zero_init (bool): If True, initializes relative positional parameters to zero.
             window_size (int): Size of attention window. If 0, uses global attention.
-            input_size (Tuple[int, int] | None): Input resolution for calculating relative positional parameter size.
+            input_size (tuple[int, int] | None): Input resolution for calculating relative positional parameter size.
 
         Examples:
             >>> block = Block(dim=256, num_heads=8, window_size=7)
@@ -1026,7 +1026,7 @@ class REAttention(nn.Module):
             qkv_bias (bool): If True, adds a learnable bias to query, key, value projections.
             use_rel_pos (bool): If True, uses relative positional encodings.
             rel_pos_zero_init (bool): If True, initializes relative positional parameters to zero.
-            input_size (Tuple[int, int] | None): Input resolution for calculating relative positional parameter size.
+            input_size (tuple[int, int] | None): Input resolution for calculating relative positional parameter size.
                 Required if use_rel_pos is True.
 
         Examples:
@@ -1106,9 +1106,9 @@ class PatchEmbed(nn.Module):
         image data into a suitable format for subsequent transformer blocks.
 
         Args:
-            kernel_size (Tuple[int, int]): Size of the convolutional kernel for patch extraction.
-            stride (Tuple[int, int]): Stride of the convolutional operation.
-            padding (Tuple[int, int]): Padding applied to the input before convolution.
+            kernel_size (tuple[int, int]): Size of the convolutional kernel for patch extraction.
+            stride (tuple[int, int]): Stride of the convolutional operation.
+            padding (tuple[int, int]): Padding applied to the input before convolution.
             in_chans (int): Number of input image channels.
             embed_dim (int): Dimensionality of the output patch embeddings.
 

ultralytics/models/sam/modules/decoders.py

@@ -329,7 +329,7 @@ class SAM2MaskDecoder(nn.Module):
             dense_prompt_embeddings (torch.Tensor): Embeddings of the mask inputs with shape (B, C, H, W).
             multimask_output (bool): Whether to return multiple masks or a single mask.
             repeat_image (bool): Flag to repeat the image embeddings.
-            high_res_features (List[torch.Tensor] | None, optional): Optional high-resolution features.
+            high_res_features (list[torch.Tensor] | None, optional): Optional high-resolution features.
 
         Returns:
             masks (torch.Tensor): Batched predicted masks with shape (B, N, H, W).