Closed ghost closed 5 years ago
While training on the GPU, setting the iterator through the GPU ID has been deprecated. Doing it this way causes all tensors to fall back to being placed on the CPU which causes undefined behavior. The device needs to be set through torch.device.
While training on the GPU, setting the iterator through the GPU ID has been deprecated. Doing it this way causes all tensors to fall back to being placed on the CPU which causes undefined behavior. The device needs to be set through torch.device.