弄清楚什么类型的 ResNet

如何解决弄清楚什么类型的 ResNet

我想弄清楚我拥有什么类型的 ResNet，但我遇到了问题。该模型可以在 here 中找到。如果您还可以解释如何解决这个问题，那就太好了。我认为它是 ResNet-18，但我只是想确定一下。我也包括以下内容。感谢您的帮助，如果这属于其他堆栈网站，请告诉我。

import torch
import torch.nn as nn
import torch.nn.functional as F

class resBlock(nn.Module): 
    def __init__(self,inchannel,outchannel,downsample=False,upsample=False):
        super(resBlock,self).__init__()

        self.inchannel = inchannel
        self.outchannel = outchannel
        self.downsample = downsample
        self.upsample = upsample

        self.reflectpad1 = nn.ReflectionPad2d(1)

        if (self.downsample == True):
            self.down = nn.Sequential(nn.ReflectionPad2d(1),nn.Conv2d(inchannel,kernel_size =3,stride=2,padding=0,bias=False),nn.BatchNorm2d(outchannel),nn.ReLU(),nn.ReflectionPad2d(1),nn.Conv2d(outchannel,stride=1,nn.BatchNorm2d(outchannel))
            self.skipDown = nn.Conv2d(inchannel,kernel_size =1,bias=False)


        if (self.downsample == False and self.upsample == False):
            self.standard = nn.Sequential(nn.ReflectionPad2d(1),nn.BatchNorm2d(outchannel))


        if (self.upsample == True):
            self.up = nn.Sequential(nn.ReflectionPad2d(1),nn.ConvTranspose2d(inchannel,padding=3,output_padding=1,nn.BatchNorm2d(outchannel))
            self.skipUp = nn.Conv2d(inchannel,bias=False)






    def forward(self,x):

        if (self.downsample == True):
            x1 = self.down(x)
            x1 += self.skipDown(x)


        if (self.downsample == False and self.upsample == False):
            x1 = self.standard(x)
            x1 += x


        if (self.upsample == True):
            x1 = self.up(x)
            x2 = self.skipUp(x)
            x1 += F.interpolate(x2,scale_factor=2)


        return F.relu(x1)



class ResNet(nn.Module):
    def __init__(self):
        super(ResNet,self).__init__()

        self.beforeRes = nn.Sequential(nn.ReflectionPad2d(3),nn.Conv2d(4,32,kernel_size =7,nn.ReflectionPad2d(2),nn.Conv2d(32,64,kernel_size =5,nn.ReLU())

        self.block0 = resBlock(64,128,downsample=True,upsample=False)

        self.block1 = nn.ModuleList([resBlock(128,upsample=False ) for i in range(6)])

        self.block2 = resBlock(128,upsample=True)

        self.afterRes = nn.Sequential(nn.ReflectionPad2d(1),nn.ConvTranspose2d(64,padding=4,nn.ReflectionPad2d(3),3,bias=True),nn.Sigmoid())


    def forward(self,x):

        x1 = self.beforeRes(x)

        x1 = self.block0(x1)

        for i in range(6):
            x1 = self.block1[i](x1)

        x1 = self.block2(x1)

        x1 = self.afterRes(x1)

        return x1