RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu

Workarounds

1. 96% success Define a single device variable and apply .to(device) to model, inputs, and labels consistently

   device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
   model = model.to(device)
   inputs = inputs.to(device)
   labels = labels.to(device)

   Sources: https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html

2. 92% success Use a DataLoader with pin_memory=True and move batches to GPU inside the training loop

   loader = DataLoader(dataset, pin_memory=True)
   for batch in loader:
       inputs, labels = batch[0].to(device), batch[1].to(device)

   Sources: https://pytorch.org/docs/stable/data.html#torch.utils.data.DataLoader

3. 88% success Check intermediate tensors created inside the model (e.g., positional encodings, masks) are also on the correct device

   Inside nn.Module.forward(), use self.weight.device or input.device to create new tensors on the correct device: mask = torch.ones(size, device=input.device)

   Sources: https://pytorch.org/docs/stable/notes/cuda.html#best-practices