model.py

from torch import nn

class model(nn.Module):
    def __init__(self, captcha_size, captcha_array):
        super(model, self).__init__()
        self.layer1=nn.Sequential(
            nn.Conv2d(in_channels=1,out_channels=64,kernel_size=3,padding=1),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2)#[6, 64, 30, 80]
        )
        self.layer2=nn.Sequential(
            nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, padding=1),
            nn.ReLU(),
            nn.MaxPool2d(2)#[6, 128, 15, 40]
        )
        self.layer3 = nn.Sequential(
            nn.Conv2d(in_channels=128, out_channels=256, kernel_size=3, padding=1),
            nn.ReLU(),
            nn.MaxPool2d(2)  # [6, 256, 7, 20]
        )
        self.layer4 = nn.Sequential(
            nn.Conv2d(in_channels=256, out_channels=512, kernel_size=3, padding=1),
            nn.ReLU(),
            nn.MaxPool2d(2)  # [6, 512, 3, 10]
        )
        self.layer6 = nn.Sequential(
            nn.Flatten(),#[6, 2560] [64, 15360]
            nn.Linear(in_features=15360,out_features=4096),
            nn.Dropout(0.2),  # drop 20% of the neuron
            nn.ReLU(),
            nn.Linear(in_features=4096,out_features=captcha_size*captcha_array.__len__())
        )
    def forward(self,x):
        x=self.layer1(x)
        x=self.layer2(x)
        x=self.layer3(x)
        x=self.layer4(x)
        x=self.layer6(x)
        return x