Ultralytics YOLO

)]): List of input images in HWC numpy array format.

     Returns:

•        (torch.Tensor): The transformed image tensor, ready for further preprocessing steps.

•        (torch.Tensor): The transformed image tensor in BCHW format, ready for further preprocessing steps.

•    Examples:

•        >>> predictor = Predictor()

•        >>> image = [np.random.rand(640, 640, 3).astype(np.float32)]

•        >>> transformed_image = predictor.pre_transform(image)

•    """
     if isinstance(im, list) and len(im):
         assert len(im) == 1, "Batched inference not supported for SAM."
         im = im[0]

  diff --git a/ultralytics/yolo/utils/ops.py b/ultralytics/yolo/utils/ops.py index fea849f4b9c..e5e62b58ed9 100644 --- a/ultralytics/yolo/utils/ops.py +++ b/ultralytics/yolo/utils/ops.py @@ -17,6 +17,7 @@ import torch.nn as nn import torch.nn.functional as F import torchvision from IPython import display # IPython.autoimport() +from typing import Tuple, Any

from ultralytics.yolo.utils import LOGGER, TryExcept, colorstr, emojis, yaml_load

@@ -24,6 +25,14 @@ from .checks import check_version

TORCH_1_13 = check_version(torch.version, "1.13.0")

+Tuple.args = (Any, Any) # Fixes Tuple deserialization + + +def get_num_params(module: nn.Module, exclude=None) -> int:

• """Calculate the total number of parameters in a module."""
• if exclude:

•    params = [p.numel() for n, p in module.named_parameters() if exclude not in n]

• return sum(p.numel() for p in module.parameters())

def try_new(op, *args, **kwargs): """Try torch.[operation](*args, **kwargs)."""