数据集下载地址：
链接：https://pan.baidu.com/s/1l1AnBgkAAEhh0vI5_loWKw
提取码：2xq4
创建数据集：https://www.cnblogs.com/xiximayou/p/12398285.html
读取数据集：https://www.cnblogs.com/xiximayou/p/12422827.html
进行训练：https://www.cnblogs.com/xiximayou/p/12448300.html
保存模型并继续进行训练：https://www.cnblogs.com/xiximayou/p/12452624.html
加载保存的模型并测试：https://www.cnblogs.com/xiximayou/p/12459499.html
划分验证集并边训练边验证：https://www.cnblogs.com/xiximayou/p/12464738.html
使用学习率衰减策略并边训练边测试：https://www.cnblogs.com/xiximayou/p/12468010.html
利用tensorboard可视化训练和测试过程：https://www.cnblogs.com/xiximayou/p/12482573.html
从命令行接收参数：https://www.cnblogs.com/xiximayou/p/12488662.html
使用top1和top5准确率来衡量模型：https://www.cnblogs.com/xiximayou/p/12489069.html
epoch、batchsize、step之间的关系：https://www.cnblogs.com/xiximayou/p/12405485.html
之前都是从头开始训练模型，本节我们要使用预训练的模型来进行训练。
只需要在train.py中加上：
if baseline: model =torchvision.models.resnet18(pretrained=False) model.fc = nn.Linear(model.fc.in_features,2,bias=False) else: print("使用预训练的resnet18模型") model=torchvision.models.resnet18(pretrained=True) for i in model.state_dict(): print(i) model.fc = nn.Linear(model.fc.in_features,2,bias=False) print(model)
使用预训练的resnet18模型 conv1.weight bn1.weight bn1.bias bn1.running_mean bn1.running_var bn1.num_batches_tracked layer1.0.conv1.weight layer1.0.bn1.weight layer1.0.bn1.bias layer1.0.bn1.running_mean layer1.0.bn1.running_var layer1.0.bn1.num_batches_tracked layer1.0.conv2.weight layer1.0.bn2.weight layer1.0.bn2.bias layer1.0.bn2.running_mean layer1.0.bn2.running_var layer1.0.bn2.num_batches_tracked layer1.1.conv1.weight layer1.1.bn1.weight layer1.1.bn1.bias layer1.1.bn1.running_mean layer1.1.bn1.running_var layer1.1.bn1.num_batches_tracked layer1.1.conv2.weight layer1.1.bn2.weight layer1.1.bn2.bias layer1.1.bn2.running_mean layer1.1.bn2.running_var layer1.1.bn2.num_batches_tracked layer2.0.conv1.weight layer2.0.bn1.weight layer2.0.bn1.bias layer2.0.bn1.running_mean layer2.0.bn1.running_var layer2.0.bn1.num_batches_tracked layer2.0.conv2.weight layer2.0.bn2.weight layer2.0.bn2.bias layer2.0.bn2.running_mean layer2.0.bn2.running_var layer2.0.bn2.num_batches_tracked layer2.0.downsample.0.weight layer2.0.downsample.1.weight layer2.0.downsample.1.bias layer2.0.downsample.1.running_mean layer2.0.downsample.1.running_var layer2.0.downsample.1.num_batches_tracked layer2.1.conv1.weight layer2.1.bn1.weight layer2.1.bn1.bias layer2.1.bn1.running_mean layer2.1.bn1.running_var layer2.1.bn1.num_batches_tracked layer2.1.conv2.weight layer2.1.bn2.weight layer2.1.bn2.bias layer2.1.bn2.running_mean layer2.1.bn2.running_var layer2.1.bn2.num_batches_tracked layer3.0.conv1.weight layer3.0.bn1.weight layer3.0.bn1.bias layer3.0.bn1.running_mean layer3.0.bn1.running_var layer3.0.bn1.num_batches_tracked layer3.0.conv2.weight layer3.0.bn2.weight layer3.0.bn2.bias layer3.0.bn2.running_mean layer3.0.bn2.running_var layer3.0.bn2.num_batches_tracked layer3.0.downsample.0.weight layer3.0.downsample.1.weight layer3.0.downsample.1.bias layer3.0.downsample.1.running_mean layer3.0.downsample.1.running_var layer3.0.downsample.1.num_batches_tracked layer3.1.conv1.weight layer3.1.bn1.weight layer3.1.bn1.bias layer3.1.bn1.running_mean layer3.1.bn1.running_var layer3.1.bn1.num_batches_tracked layer3.1.conv2.weight layer3.1.bn2.weight layer3.1.bn2.bias layer3.1.bn2.running_mean layer3.1.bn2.running_var layer3.1.bn2.num_batches_tracked layer4.0.conv1.weight layer4.0.bn1.weight layer4.0.bn1.bias layer4.0.bn1.running_mean layer4.0.bn1.running_var layer4.0.bn1.num_batches_tracked layer4.0.conv2.weight layer4.0.bn2.weight layer4.0.bn2.bias layer4.0.bn2.running_mean layer4.0.bn2.running_var layer4.0.bn2.num_batches_tracked layer4.0.downsample.0.weight layer4.0.downsample.1.weight layer4.0.downsample.1.bias layer4.0.downsample.1.running_mean layer4.0.downsample.1.running_var layer4.0.downsample.1.num_batches_tracked layer4.1.conv1.weight layer4.1.bn1.weight layer4.1.bn1.bias layer4.1.bn1.running_mean layer4.1.bn1.running_var layer4.1.bn1.num_batches_tracked layer4.1.conv2.weight layer4.1.bn2.weight layer4.1.bn2.bias layer4.1.bn2.running_mean layer4.1.bn2.running_var layer4.1.bn2.num_batches_tracked fc.weight fc.bias ResNet( (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False) (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False) (layer1): Sequential( (0): BasicBlock( (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) (1): BasicBlock( (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (layer2): Sequential( (0): BasicBlock( (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False) (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (downsample): Sequential( (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False) (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (1): BasicBlock( (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (layer3): Sequential( (0): BasicBlock( (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False) (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (downsample): Sequential( (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False) (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (1): BasicBlock( (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (layer4): Sequential( (0): BasicBlock( (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False) (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (downsample): Sequential( (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False) (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (1): BasicBlock( (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (avgpool): AdaptiveAvgPool2d(output_size=(1, 1)) (fc): Linear(in_features=512, out_features=2, bias=False) )
接下来来看看如何冻结某些层，不让其在训练的时候进行梯度更新。
首先我们输出下信息看看结构:
i=0
for child in model.children():
i+=1
print("第{}个child".format(str(i))) print(child)
第1个child Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False) 第2个child BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) 第3个child ReLU(inplace=True) 第4个child MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False) 第5个child Sequential( (0): BasicBlock( (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) (1): BasicBlock( (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) 第6个child Sequential( (0): BasicBlock( (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False) (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (downsample): Sequential( (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False) (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (1): BasicBlock( (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) 第7个child Sequential( (0): BasicBlock( (conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False) (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (downsample): Sequential( (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False) (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (1): BasicBlock( (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) 第8个child Sequential( (0): BasicBlock( (conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False) (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (downsample): Sequential( (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False) (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) (1): BasicBlock( (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): ReLU(inplace=True) (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) ) ) 第9个child AdaptiveAvgPool2d(output_size=(1, 1)) 第10个child Linear(in_features=512, out_features=2, bias=False)
我们冻结前面的7个child，只更新第8、9、10个child的参数。可这么定义：
print("使用预训练的resnet18模型") model=torchvision.models.resnet18(pretrained=True) model.fc = nn.Linear(model.fc.in_features,2,bias=False) i=0 for child in model.children(): i+=1 #print("第{}个child".format(str(i))) #print(child) if i<=7: for param in child.parameters(): param.requires_grad=False #我们打印下是否是设置成功 for name, param in model.named_parameters(): if param.requires_grad: print("需要梯度：", name) else: print("不需要梯度：", name)
接下来我们还要在优化器中过滤掉不需要更新参数的层：
optimizer = torch.optim.SGD(filter(lambda p: p.requires_grad, model.parameters()), lr=0.1, momentum=0.9, weight_decay=1*1e-4)
结果：
使用预训练的resnet18模型 不需要梯度： conv1.weight 不需要梯度： bn1.weight 不需要梯度： bn1.bias 不需要梯度： layer1.0.conv1.weight 不需要梯度： layer1.0.bn1.weight 不需要梯度： layer1.0.bn1.bias 不需要梯度： layer1.0.conv2.weight 不需要梯度： layer1.0.bn2.weight 不需要梯度： layer1.0.bn2.bias 不需要梯度： layer1.1.conv1.weight 不需要梯度： layer1.1.bn1.weight 不需要梯度： layer1.1.bn1.bias 不需要梯度： layer1.1.conv2.weight 不需要梯度： layer1.1.bn2.weight 不需要梯度： layer1.1.bn2.bias 不需要梯度： layer2.0.conv1.weight 不需要梯度： layer2.0.bn1.weight 不需要梯度： layer2.0.bn1.bias 不需要梯度： layer2.0.conv2.weight 不需要梯度： layer2.0.bn2.weight 不需要梯度： layer2.0.bn2.bias 不需要梯度： layer2.0.downsample.0.weight 不需要梯度： layer2.0.downsample.1.weight 不需要梯度： layer2.0.downsample.1.bias 不需要梯度： layer2.1.conv1.weight 不需要梯度： layer2.1.bn1.weight 不需要梯度： layer2.1.bn1.bias 不需要梯度： layer2.1.conv2.weight 不需要梯度： layer2.1.bn2.weight 不需要梯度： layer2.1.bn2.bias 不需要梯度： layer3.0.conv1.weight 不需要梯度： layer3.0.bn1.weight 不需要梯度： layer3.0.bn1.bias 不需要梯度： layer3.0.conv2.weight 不需要梯度： layer3.0.bn2.weight 不需要梯度： layer3.0.bn2.bias 不需要梯度： layer3.0.downsample.0.weight 不需要梯度： layer3.0.downsample.1.weight 不需要梯度： layer3.0.downsample.1.bias 不需要梯度： layer3.1.conv1.weight 不需要梯度： layer3.1.bn1.weight 不需要梯度： layer3.1.bn1.bias 不需要梯度： layer3.1.conv2.weight 不需要梯度： layer3.1.bn2.weight 不需要梯度： layer3.1.bn2.bias 需要梯度： layer4.0.conv1.weight 需要梯度： layer4.0.bn1.weight 需要梯度： layer4.0.bn1.bias 需要梯度： layer4.0.conv2.weight 需要梯度： layer4.0.bn2.weight 需要梯度： layer4.0.bn2.bias 需要梯度： layer4.0.downsample.0.weight 需要梯度： layer4.0.downsample.1.weight 需要梯度： layer4.0.downsample.1.bias 需要梯度： layer4.1.conv1.weight 需要梯度： layer4.1.bn1.weight 需要梯度： layer4.1.bn1.bias 需要梯度： layer4.1.conv2.weight 需要梯度： layer4.1.bn2.weight 需要梯度： layer4.1.bn2.bias 需要梯度： fc.weight
拓展：如果是我们自己定义的模型和预训练的模型不一致应该怎么加载参数呢？
这里以以resnet50为例，这里我们再新定义一个卷积神经网络：
# coding=UTF-8 import torchvision.models as models import torch import torch.nn as nn import math import torch.utils.model_zoo as model_zoo class CNN(nn.Module): def __init__(self, block, layers, num_classes=2): self.inplanes = 64 super(ResNet, self).__init__() self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.BatchNorm2d(64) self.relu = nn.ReLU(inplace=True) self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) self.layer1 = self._make_layer(block, 64, layers[0]) self.layer2 = self._make_layer(block, 128, layers[1], stride=2) self.layer3 = self._make_layer(block, 256, layers[2], stride=2) self.layer4 = self._make_layer(block, 512, layers[3], stride=2) self.avgpool = nn.AvgPool2d(7, stride=1) #新增一个反卷积层 self.convtranspose1 = nn.ConvTranspose2d(2048, 2048, kernel_size=3, stride=1, padding=1, output_padding=0, groups=1, bias=False, dilation=1) #新增一个最大池化层 self.maxpool2 = nn.MaxPool2d(kernel_size=3, stride=1, padding=1) #去掉原来的fc层，新增一个fclass层 self.fclass = nn.Linear(2048, num_classes) for m in self.modules(): if isinstance(m, nn.Conv2d): n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels m.weight.data.normal_(0, math.sqrt(2. / n)) elif isinstance(m, nn.BatchNorm2d): m.weight.data.fill_(1) m.bias.data.zero_() def _make_layer(self, block, planes, blocks, stride=1): downsample = None if stride != 1 or self.inplanes != planes * block.expansion: downsample = nn.Sequential( nn.Conv2d(self.inplanes, planes * block.expansion, kernel_size=1, stride=stride, bias=False), nn.BatchNorm2d(planes * block.expansion), ) layers = [] layers.append(block(self.inplanes, planes, stride, downsample)) self.inplanes = planes * block.expansion for i in range(1, blocks): layers.append(block(self.inplanes, planes)) return nn.Sequential(*layers) def forward(self, x): x = self.conv1(x) x = self.bn1(x) x = self.relu(x) x = self.maxpool(x) x = self.layer1(x) x = self.layer2(x) x = self.layer3(x) x = self.layer4(x) x = self.avgpool(x) #新加层的forward x = x.view(x.size(0), -1) x = self.convtranspose1(x) x = self.maxpool2(x) x = x.view(x.size(0), -1) x = self.fclass(x) return x #加载model resnet50 = models.resnet50(pretrained=True) cnn = CNN(Bottleneck, [3, 4, 6, 3]) #读取参数
#取出预训练模型的参数 pretrained_dict = resnet50.state_dict()
#取出本模型的参数 model_dict = cnn.state_dict() # 将pretrained_dict里不属于model_dict的键剔除掉 pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict} # 更新现有的model_dict model_dict.update(pretrained_dict) # 加载我们真正需要的state_dict cnn.load_state_dict(model_dict) # print(resnet50) print(cnn)
下面也摘取了一些使用部分预训练模型初始化网络的方法：
方式一： 自己网络和预训练网络结构一致的层，使用预训练网络对应层的参数批量初始化
model_dict = model.state_dict() # 取出自己网络的参数字典 pretrained_dict = torch.load("I:/迅雷下载/alexnet-owt-4df8aa71.pth")# 加载预训练网络的参数字典 # 取出预训练网络的参数字典 keys = [] for k, v in pretrained_dict.items(): keys.append(k) i = 0 # 自己网络和预训练网络结构一致的层，使用预训练网络对应层的参数初始化 for k, v in model_dict.items(): if v.size() == pretrained_dict[keys[i]].size(): model_dict[k] = pretrained_dict[keys[i]] #print(model_dict[k]) i = i + 1 model.load_state_dict(model_dict)
方式二：自己网络和预训练网络结构一致的层，按层初始化
# 加粗自己定义一个网络叫CNN model = CNN（） model_dict = model.state_dict() # 取出自己网络的参数 for k, v in model_dict.items(): # 查看自己网络参数各层叫什么名称 print(k) pretrained_dict = torch.load("I:/迅雷下载/alexnet-owt-4df8aa71.pth")# 加载预训练网络的参数 for k, v in pretrained_dict.items(): # 查看预训练网络参数各层叫什么名称 print(k) # 对应层赋值初始化 model_dict['conv1.0.weight'] = pretrained_dict['features.0.weight'] # 将自己网络的conv1.0层的权重初始化为预训练网络features.0层的权重 model_dict['conv1.0.bias'] = pretrained_dict['features.0.bias'] # 将自己网络的conv1.0层的偏置项初始化为预训练网络features.0层的偏置项 model_dict['conv2.1.weight'] = pretrained_dict['features.3.weight'] model_dict['conv1.1.bias'] = pretrained_dict['features.3.bias'] model_dict['conv2.1.weight'] = pretrained_dict['features.6.weight'] model_dict['conv2.1.bias'] = pretrained_dict['features.6.bias'] ... ...
下一节补充下计算数据集的标准差和方差，在数据增强时对数据进行标准化的时候用。
参考：
https://blog.csdn.net/feizai1208917009/article/details/103598233