please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU

Question

# set the computation device
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Load model checkpoint
checkpoint = 'checkpoints/checkpoint_ssd300.pth.tar'
checkpoint = torch.load(checkpoint)
start_epoch = checkpoint['epoch'] + 1
print('\nLoaded checkpoint from epoch %d.\n' % start_epoch)
model = checkpoint['model']
model = model.to(device)
model.eval()

When I try to run this code block, I get the following problem:

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

Answer 1

The error message is indicating that you are trying to load a model checkpoint that was trained on a GPU (CUDA device), but your current machine does not have a GPU or CUDA is not available.

The line device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') is trying to determine whether CUDA is available on the current machine, and if it is, it sets the device variable to 'cuda', otherwise it sets it to 'cpu'.

The line checkpoint = torch.load(checkpoint) is trying to load the model checkpoint from the specified file, but it is trying to do so on the 'cuda' device, which is causing the error.

To resolve this issue, you can use the map_location argument of the torch.load function to specify that the model should be loaded on the 'cpu' device, instead of the 'cuda' device.

checkpoint = torch.load(checkpoint, map_location=torch.device('cpu'))

This way the model will be loaded on the CPU device, even if a CUDA device was used to train it.