evaluations/fidkid-pytorch/models/lenet.py

import torch.nn as nn
from collections import OrderedDict


class LeNet5(nn.Module):
    """
    Input - 1x32x32
    C1 - 6@28x28 (5x5 kernel)
    tanh
    S2 - 6@14x14 (2x2 kernel, stride 2) Subsampling
    C3 - 16@10x10 (5x5 kernel, complicated shit)
    tanh
    S4 - 16@5x5 (2x2 kernel, stride 2) Subsampling
    C5 - 120@1x1 (5x5 kernel)
    F6 - 84
    tanh
    F7 - 10 (Output)
    """
    def __init__(self):
        super(LeNet5, self).__init__()

        self.convnet = nn.Sequential(OrderedDict([
            ('c1', nn.Conv2d(1, 6, kernel_size=(5, 5))),
            ('tanh1', nn.Tanh()),
            ('s2', nn.MaxPool2d(kernel_size=(2, 2), stride=2, padding=1)),
            ('c3', nn.Conv2d(6, 16, kernel_size=(5, 5))),
            ('tanh3', nn.Tanh()),
            ('s4', nn.MaxPool2d(kernel_size=(2, 2), stride=2, padding=1)),
            ('c5', nn.Conv2d(16, 120, kernel_size=(5, 5))),
            ('tanh5', nn.Tanh())
        ]))

        self.fc = nn.Sequential(OrderedDict([
            ('f6', nn.Linear(120, 84)),
            ('tanh6', nn.Tanh()),
            ('f7', nn.Linear(84, 10)),
            ('sig7', nn.LogSoftmax(dim=-1))
        ]))

    def forward(self, img):
        output = self.convnet(img)
        output = output.view(img.size(0), -1)
        output = self.fc(output)
        return output

    def extract_features(self, img):
        output = self.convnet(img.float())
        output = output.view(img.size(0), -1)
        output = self.fc[1](self.fc[0](output))
        return output