首先安装torchsummary
pip3 install torchsummary
执行如下代码:
from torchvision import models
model = models.vgg16()
torchsummary.summary(model, (3, 224, 224))
报错:argument 0 is not a Variable
解决方法是:pytorch版本太老
更新版本:pip3 install --upgrade torch
再次执行又报错:RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same
解决方法是模型在cuda上了:
from torchvision import models
model = models.vgg16()
torchsummary.summary(model.cuda(), (3, 224, 224))
安装完torch1.0之后,在import torch时报错:
ImportError: libcurand.so.9.1: cannot open shared object file: No such file or directory
解决办法:执行如下命令
unset LD_LIBRARY_PATH