Why does torch.zeroes(10,10,10).to("cuda" or "gpu") return a tensor full of nans?

yea why does to.(DEVICE) with DEVICE = “gpu” or "cuda return a tensor full of nans?

"gpu" shouldn’t be a valid argument to the .to() operation.
Could you please post a small code snippet to reproduce this issue and also post information about your setup, i.e. PyTorch, CUDA, version, NVIDIA driver, etc.?