resnet.py

import torch
import torch.nn as nn


class ResNet(nn.Module):
    def __init__(self, in_channels: int, num_classes: int):
        """ResNet9"""
        super().__init__()
        
        self.conv1 = ConvBlock(in_channels, 64)
        self.conv2 = ConvBlock(64, 128, pool=True)
        self.res1 = nn.Sequential(
            ConvBlock(128, 128),
            ConvBlock(128, 128)
        )

        self.conv3 = ConvBlock(128, 256)
        self.conv4 = ConvBlock(256, 512, pool=True)
        self.res2 = nn.Sequential(
            ConvBlock(512, 512),
            ConvBlock(512, 512)
        )
        
        self.classifier = nn.Sequential(
            nn.MaxPool2d(kernel_size=(4, 4)),
            nn.AdaptiveAvgPool2d(1),
            nn.Flatten(), 
            nn.Linear(512, 128),
            nn.Dropout(0.25),
            nn.Linear(128, num_classes),
            nn.Dropout(0.25),
        )
        
    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.conv1(x)
        x = self.conv2(x)
        x = self.res1(x) + x #skip
        x = self.conv3(x)
        x = self.conv4(x)
        x = self.res2(x) + x #skip
        prediction = self.classifier(x)
        return prediction

class ConvBlock(nn.Module):
    def __init__(self, in_channels: int, out_channels: int, pool: bool = False, pool_no: int = 2):
        super().__init__()
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.pool = pool
        self.pool_no = pool_no

        if self.pool:
            self.pool_block = nn.Sequential(
                nn.ReLU(inplace=True),
                nn.MaxPool2d(self.pool_no)
            )
        else:
            self.pool_block = nn.Sequential(
                nn.ReLU(inplace=True),
            )
        
        self.block = nn.Sequential(
            nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1),
            nn.BatchNorm2d(out_channels),
            self.pool_block
        )
    
    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.block(x)
        return x