Pytorch 报错总结

目前在学习pytorch，自己写了一些例子，在这里记录下来一些报错及总结

1. RuntimeError: Expected object of type torch.FloatTensor but found type torch.cuda.FloatTensor for argument #2 'weight'

详细报错信息

 Traceback (most recent call last):
   File "dogvscat-resnet.py", line , in <module>
     outputs = net(inputs)
   File "/home/lzx/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py", line , in __call__
     result = self.forward(*input, **kwargs)
   File "/home/lzx/anaconda3/envs/pytorch/lib/python3.6/site-packages/torchvision-0.2.1-py3.6.egg/torchvision/models/resnet.py", li
 ne , in forward
   File "/home/lzx/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py", line , in __call__
     result = self.forward(*input, **kwargs)
   File "/home/lzx/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/modules/conv.py", line , in forward
     self.padding, self.dilation, self.groups)
 RuntimeError: Expected object of type torch.FloatTensor but found type torch.cuda.FloatTensor for argument # 'weight'

参考：https://github.com/wohlert/semi-supervised-pytorch/issues/7

这个报错其实比较隐蔽，用Google搜索的第一页都没什么参考价值，只有上面的这个链接里提醒了我，

在GPU上进行训练时，需要把模型和数据都加上.cuda()，如

model.cuda()

但是对于数据，这个.cuda()并非是inplace操作，就是说不单单是在变量名后面加上.cuda()就可以了

还必须显示的赋值回去，即：

data.cuda()是不行的，而

data = data.cuda()才是可以的。

这样的显示声明的细节非常重要。

示例代码：用LeNet做猫狗的二分类，自己写的代码

请重点关注以下行的写法：46 47 57 58 96 97

 import os
 from PIL import Image
 import numpy as np
 import torch
 from torchvision import transforms as T
 from torchvision.datasets import ImageFolder
 from torch.utils.data import DataLoader
 import torch.nn as nn
 import torch.nn.functional as F
 from torch import optim
 from torch.utils import data
 import torchvision as tv
 from torchvision.transforms import ToPILImage
 show = ToPILImage()  # 可以把Tensor转成Image，方便可视化

 transform = T.Compose([
     T.Resize(32), # 缩放图片(Image)，保持长宽比不变，最短边为224像素
     T.CenterCrop(32), # 从图片中间切出224*224的图片
     T.ToTensor(), # 将图片(Image)转成Tensor，归一化至[0, 1]
     T.Normalize(mean=[.5, .5, .5], std=[.5, .5, .5]) # 标准化至[-1, 1]，规定均值和标准差
 ])

 class Net(nn.Module):
     def __init__(self):
         super(Net, self).__init__()
         self.conv1 = nn.Conv2d(3, 6, 5)
         self.conv2 = nn.Conv2d(6, 16, 5)
         self.fc1 = nn.Linear(16*5*5, 120)
         self.fc2 = nn.Linear(120, 84)
         self.fc3 = nn.Linear(84, 2)

     def forward(self, x):
         x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
         x = F.max_pool2d(F.relu(self.conv2(x)), 2)
         x = x.view(x.size()[0], -1)
         x = F.relu(self.fc1(x))
         x = F.relu(self.fc2(x))
         x = self.fc3(x)
         return x

 net = Net()
 if torch.cuda.is_available():
     print("Using GPU")
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 net.to(device)

 def test():
     correct = 0 # 预测正确的图片数
     total = 0 # 总共的图片数
     # 由于测试的时候不需要求导，可以暂时关闭autograd，提高速度，节约内存
     with torch.no_grad():
         for data in testloader:
             images, labels = data
             images = images.to(device)
             labels = labels.to(device)
             outputs = net(images)
             _, predicted = torch.max(outputs, 1)
             total += labels.size(0)
             correct += (predicted == labels).sum()

         print('Accuracy in the test dataset: %.1f %%' % (100 * correct / total))

 train_dataset = ImageFolder('/home/lzx/datasets/dogcat/sub-train/', transform=transform)
 test_dataset = ImageFolder('/home/lzx/datasets/dogcat/sub-test/', transform=transform)
 # dataset = DogCat('/home/lzx/datasets/dogcat/sub-train/', transforms=transform)
 # train_dataset = ImageFolder('/Users/lizhixuan/PycharmProjects/pytorch_learning/Chapter5/sub-train/', transform=transform)
 # test_dataset = ImageFolder('/Users/lizhixuan/PycharmProjects/pytorch_learning/Chapter5/sub-test/', transform=transform)

 trainloader = torch.utils.data.DataLoader(
                     train_dataset,
                     batch_size=512,
                     shuffle=True,
                     num_workers=4)
 testloader = torch.utils.data.DataLoader(
                     test_dataset,
                     batch_size=512,
                     shuffle=False,
                     num_workers=4)
 classes = ('cat', 'dog')

 criterion = nn.CrossEntropyLoss()  # 交叉熵损失函数
 optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)

 print("Starting to train")
 torch.set_num_threads(8)
 for epoch in range(1000):

     running_loss = 0.0
     for i, data in enumerate(trainloader, 0):

         # 输入数据
         inputs, labels = data
         inputs = inputs.to(device)
         labels = labels.to(device)

         # 梯度清零
         optimizer.zero_grad()

         # forward + backward
         outputs = net(inputs)
         loss = criterion(outputs, labels)
 #         print("outputs %s  labels %s" % (outputs, labels))
         loss.backward()

         # 更新参数
         optimizer.step()

         # 打印log信息
         # loss 是一个scalar,需要使用loss.item()来获取数值，不能使用loss[0]
         running_loss += loss.item()
         print_gap = 10
         if i % print_gap == (print_gap-1): # 每1000个batch打印一下训练状态
             print('[%d, %5d] loss: %.3f' \
                   % (epoch+1, i+1, running_loss / print_gap))
             running_loss = 0.0
     test()
 print('Finished Training')

这样一来，就完全明白了如何把代码放在GPU上运行了，哈哈