classifier

Classifier

Bases: nn.Module

A 3-layer MLP for classification.

Source code in pytorch_adapt\models\classifier.py

class Classifier(nn.Module):
    """
    A 3-layer MLP for classification.
    """

    def __init__(self, num_classes, in_size=2048, h=1024):
        """
        Arguments:
            num_classes: size of the output
            in_size: size of the input
            h: hidden layer size
        """
        super().__init__()
        self.h = h
        self.net = nn.Sequential(
            nn.Linear(in_size, h),
            nn.ReLU(),
            nn.Dropout(),
            nn.Linear(h, h // 2),
            nn.ReLU(),
            nn.Dropout(),
            nn.Linear(h // 2, num_classes),
        )

    def forward(self, x):
        """"""
        return self.net(x)

__init__(num_classes, in_size=2048, h=1024)

Parameters:

Name	Type	Description	Default
num_classes		size of the output	required
in_size		size of the input	2048
h		hidden layer size	1024

Source code in pytorch_adapt\models\classifier.py

def __init__(self, num_classes, in_size=2048, h=1024):
    """
    Arguments:
        num_classes: size of the output
        in_size: size of the input
        h: hidden layer size
    """
    super().__init__()
    self.h = h
    self.net = nn.Sequential(
        nn.Linear(in_size, h),
        nn.ReLU(),
        nn.Dropout(),
        nn.Linear(h, h // 2),
        nn.ReLU(),
        nn.Dropout(),
        nn.Linear(h // 2, num_classes),
    )