pytorch-torchvision

  目前，models模块支持Alexnet，VGG，resnet，squeezeNet，denseNet和inception V3这几个模型。可以直接构建一个网络结构（参数随机化）：

import torchvision.models as models

resnet18 = models.resnet18()
alexnet = models.alexnet()
vgg16 = models.vgg16()
squeezenet = models.squeezenet1_0()
densenet = models.densenet161()
inception = models.inception_v3()

  如果想得到预训练后的网络，就把pretrained这个参数置成true：

import torchvision.models as models
resnet18 = models.resnet18(pretrained=True)

以VGG网络模型为例，有以下结构：

torchvision.models.vgg11(pretrained=False, **kwargs)
torchvision.models.vgg11_bn(pretrained=False, **kwargs)
torchvision.models.vgg13(pretrained=False, **kwargs)
torchvision.models.vgg13_bn(pretrained=False, **kwargs)
torchvision.models.vgg16(pretrained=False, **kwargs)
torchvision.models.vgg16_bn(pretrained=False, **kwargs)
torchvision.models.vgg19(pretrained=False, **kwargs)
torchvision.models.vgg19_bn(pretrained=False, **kwargs)