Skip to content

unitorch.models.swin¤

SwinProcessor¤

Bases: HfImageClassificationProcessor

Initializes a SwinProcessor for image classification tasks.

Parameters:

Name Type Description Default
vision_config_path str

The path to the ViTImageProcessor configuration file.

 required
Source code in src/unitorch/models/swin/processing.py 
17
18
19
20
21
22
23
24
25
26
27
28
29
30
def __init__(
    self,
    vision_config_path: str,
):
    """
    Initializes a SwinProcessor for image classification tasks.

    Args:
        vision_config_path (str): The path to the ViTImageProcessor configuration file.
    """
    vision_processor = ViTImageProcessor.from_json_file(vision_config_path)
    super().__init__(
        vision_processor=vision_processor,
    )

SwinForImageClassification¤

Bases: GenericModel

Initializes a SwinForImageClassification model for image classification tasks.

Parameters:

Name Type Description Default
config_path str

The path to the Swin Transformer configuration file.

 required
num_classes int

The number of classes for classification. Defaults to 1.

 1
Source code in src/unitorch/models/swin/modeling.py 
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
def __init__(
    self,
    config_path: str,
    num_classes: Optional[int] = 1,
):
    """
    Initializes a SwinForImageClassification model for image classification tasks.

    Args:
        config_path (str): The path to the Swin Transformer configuration file.
        num_classes (int, optional): The number of classes for classification. Defaults to 1.
    """
    super().__init__()
    config = SwinConfig.from_json_file(config_path)

    self.swin = SwinModel(config)
    self.classifier = nn.Linear(self.swin.num_features, num_classes)
    self.init_weights()

forward ¤

forward(pixel_values: Tensor)

Performs a forward pass of the SwinForImageClassification model.

Parameters:

Name Type Description Default
pixel_values Tensor

The input pixel values of the image.

 required

Returns:

Type Description
Tensor

The model's logits.
Source code in src/unitorch/models/swin/modeling.py 
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
def forward(
    self,
    pixel_values: torch.Tensor,
):
    """
    Performs a forward pass of the SwinForImageClassification model.

    Args:
        pixel_values (torch.Tensor): The input pixel values of the image.

    Returns:
        (torch.Tensor):The model's logits.
    """
    outputs = self.swin(
        pixel_values=pixel_values,
    )

    pooled_output = outputs[1]
    return self.classifier(pooled_output)