pytorch教程及实例

pytorch的语法，参考博客：https://www.cnblogs.com/hellcat/tag/PyTorch/default.html?page=1

一、Resnet——pytorch模型的快速实现

import torch as t
import torch.nn as nn
from torch.nn import functional as F
 
 
class ResidualBlock(nn.Module):
    def __init__(self, inchannel, outchannel, stride=1, shortcut=None):
        super(ResidualBlock, self).__init__()
        self.left = nn.Sequential(
            nn.Conv2d(inchannel, outchannel, kernel_size=3, stride=stride, padding=1),
            nn.BatchNorm2d(outchannel),
            nn.ReLU(inplace=True),
            nn.Conv2d(outchannel, outchannel, 3, 1, 1, bias=False),
            nn.BatchNorm2d(outchannel)
        )
        self.right = shortcut
 
    def forward(self, x):
        out = self.left(x)
        residual = x if self.right is None else self.right(x)
        out += residual
        return F.relu(out)
 
 
class ResNet(nn.Module):
    def __init__(self, num_classes=1000):
        super(ResNet, self).__init__()
        self.pre = nn.Sequential(
            nn.Conv2d(3, 64, 7, 2, 3, bias=False),
            nn.BatchNorm2d(64),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        )
        self.layer1 = self._make_layer(inchannel=64, outchannel=128, block_num=3)
        self.layer2 = self._make_layer(inchannel=128, outchannel=256, block_num=4, stride=2)
        self.layer3 = self._make_layer(inchannel=256, outchannel=512, block_num=6, stride=2)
        self.layer4 = self._make_layer(inchannel=512, outchannel=512, block_num=3, stride=2)
 
        self.fc = nn.Linear(512, num_classes)
 
    def _make_layer(self, inchannel, outchannel, block_num, stride=1):
        shortcut = nn.Sequential(
            nn.Conv2d(inchannel, outchannel, 1, stride, bias=False),
            nn.BatchNorm2d(outchannel)
        )
        layers = []
        layers.append(ResidualBlock(inchannel, outchannel, stride, shortcut))
        for i in range(1, block_num):
            layers.append(ResidualBlock(outchannel, outchannel))
        return nn.Sequential(*layers)
 
    def forward(self, x):
        x = self.pre(x)  # [1, 64, 56, 56]
        x = self.layer1(x)  # [1, 128, 56, 56]
        x = self.layer2(x)  # [1, 256, 28, 28]
        x = self.layer3(x)  # [1, 512, 14, 14]
        x = self.layer4(x)  # [1, 512, 7, 7]
 
 
 
        x = F.avg_pool2d(x, 7)
        x = x.view(x.size(0), -1)
        return self.fc(x)
 
def hook(module, inputdata, output):
    '''把这层的输出拷贝到features中'''
    print("钩子输出：", output.data.size())
 
 
module = ResNet()
img = t.autograd.Variable(t.randn(1, 3, 224, 224))
handle = module.pre[0].register_forward_hook(hook)
out = module(img)
handle.remove()
print(out)

上面代码中，我们注册了钩子尝试分析一下中间的输出，可以看到，torch中的卷积层默认是SAME模式，输出就是in/stride，和TensorFlow一致。

结果：