RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same

Question

I am trying to train the following CNN as follows, but I keep getting the same error regarding .cuda() and I am not sure how to fix it. Here is a chunk of my code so far.

import matplotlib.pyplot as plt import numpy as np import torch from torch import nn from torch import optim import torch.nn.functional as F import torchvision from torchvision import datasets, transforms, models from torch.utils.data.sampler import SubsetRandomSampler   data_dir = "/home/ubuntu/ML2/ExamII/train2/" valid_size = .2  # Normalize the test and train sets with torchvision train_transforms = transforms.Compose([transforms.Resize(224),                                            transforms.ToTensor(),                                            ])  test_transforms = transforms.Compose([transforms.Resize(224),                                           transforms.ToTensor(),                                           ])  # ImageFolder class to load the train and test images train_data = datasets.ImageFolder(data_dir, transform=train_transforms) test_data = datasets.ImageFolder(data_dir, transform=test_transforms)   # Number of train images num_train = len(train_data) indices = list(range(num_train)) # Split = 20% of train images split = int(np.floor(valid_size * num_train)) # Shuffle indices of train images np.random.shuffle(indices) # Subset indices for test and train train_idx, test_idx = indices[split:], indices[:split] # Samples elements randomly from a given list of indices train_sampler = SubsetRandomSampler(train_idx) test_sampler = SubsetRandomSampler(test_idx) # Batch and load the images trainloader = torch.utils.data.DataLoader(train_data, sampler=train_sampler, batch_size=1) testloader = torch.utils.data.DataLoader(test_data, sampler=test_sampler, batch_size=1)   #print(trainloader.dataset.classes)  device = torch.device("cuda" if torch.cuda.is_available() else "cpu") model = models.resnet50(pretrained=True)  model.fc = nn.Sequential(nn.Linear(2048, 512),                                  nn.ReLU(),                                  nn.Dropout(0.2),                                  nn.Linear(512, 10),                                  nn.LogSigmoid())                                  # nn.LogSoftmax(dim=1)) # criterion = nn.NLLLoss() criterion = nn.BCELoss() optimizer = optim.Adam(model.fc.parameters(), lr=0.003) model.to(device)  #Train the network for epoch in range(2):  # loop over the dataset multiple times      running_loss = 0.0     for i, data in enumerate(trainloader, 0):         # get the inputs; data is a list of [inputs, labels]         inputs, labels = data          # zero the parameter gradients         optimizer.zero_grad()          # forward + backward + optimize         outputs = model(inputs)         loss = criterion(outputs, labels)         loss.backward()         optimizer.step()          # print statistics         running_loss += loss.item()         if i % 2000 == 1999:    # print every 2000 mini-batches             print('[%d, %5d] loss: %.3f' %                   (epoch + 1, i + 1, running_loss / 2000))             running_loss = 0.0  print('Finished Training') 

However, I keep getting this error in the console:

RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same`

Any thoughts on how to fix it? I read that maybe the model hasn't been pushed into my GPU, but not sure how to fix it. Thanks!

Answer 1

You get this error because your model is on the GPU, but your data is on the CPU. So, you need to send your input tensors to the GPU.

inputs, labels = data                         # this is what you had inputs, labels = inputs.cuda(), labels.cuda() # add this line 

Or like this, to stay consistent with the rest of your code:

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")  inputs, labels = inputs.to(device), labels.to(device) 

The same error will be raised if your input tensors are on the GPU but your model weights aren't. In this case, you need to send your model weights to the GPU.

model = MyModel()  if torch.cuda.is_available():     model.cuda() 

Here is the documentation for cuda() and cpu(), its opposite.

Answer 2

The new API is to use .to() method.

The advantage is obvious and important. Your device may tomorrow be something other than "cuda":

• cpu
• cuda
• mkldnn
• opengl
• opencl
• ideep
• hip
• msnpu
• xla

So try to avoid model.cuda() It is not wrong to check for the device

dev = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu") 

or to hardcode it:

dev=torch.device("cuda")  

same as:

dev="cuda" 

In general you can use this code:

model.to(dev) data = data.to(dev)