Week09 Day43

import torch.nn as nn


class BasicBlock(nn.Module):
    def __init__(self, in_channel: int, out_channels: [int, int]):
        super(BasicBlock, self).__init__()
        out1, out2 = out_channels
        self.conv1 = nn.Conv2d(in_channel, out1, 1, 1)
        self.relu1 = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(out1, out2, 3, 1, 1)
        self.relu2 = nn.ReLU(inplace=True)

    def forward(self, x):
        x = self.conv1(x)
        x = self.relu1(x)
        x = self.conv2(x)
        return self.relu2(x)


class DailyMission(nn.Module):
    def __init__(self, in_channel):
        super(DailyMission, self).__init__()
        self.conv1 = nn.Conv2d(in_channel, 32, 3, 1, 1)
        self.relu1 = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(32, 64, 3, 2, 1)
        self.relu2 = nn.ReLU(inplace=True)
        self.block1 = self.multiple_blocks(64, (32, 64), 1)
        self.conv3 = nn.Conv2d(64, 128, 3, 2, 1)
        self.relu3 = nn.ReLU(inplace=True)
        self.block2 = self.multiple_blocks(128, (64, 128), 2)
        self.conv4 = nn.Conv2d(128, 256, 3, 2, 1)
        self.relu4 = nn.ReLU(inplace=True)
        self.block3 = self.multiple_blocks(256, (128, 256), 8)
        self.conv5 = nn.Conv2d(256, 512, 3, 2, 1)
        self.relu5 = nn.ReLU(inplace=True)
        self.block4 = self.multiple_blocks(512, (256, 512), 8)
        self.conv6 = nn.Conv2d(512, 1024, 3, 2, 1)
        self.relu6 = nn.ReLU(inplace=True)
        self.block5 = self.multiple_blocks(1024, (512, 1024), 4)
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.fc = nn.Linear(1024, 1000)

    def forward(self, x):
        features = nn.Sequential(
            self.conv1, self.relu1,
            self.conv2, self.relu2,
            self.block1,
            self.conv3, self.relu3,
            self.block2,
            self.conv4, self.relu4,
            self.block3,
            self.conv5, self.relu5,
            self.block4,
            self.conv6, self.relu6,
            self.block5,
            self.avg_pool
        )
        x = features(x).squeeze()
        return self.fc(x)

    def multiple_blocks(self, in_channel, out_channels, iter_nums):
        blocks = []
        out1, out2 = out_channels
        for _ in range(iter_nums):
            blocks.append(BasicBlock(in_channel, (out1, out2)))
            in_channel = out2
        return nn.Sequential(*blocks)