torch.Tensor.pin_memory

Tensor.pin_memory() → Tensor

Copies the tensor to pinned memory, if it’s not already pinned. By default, the device pinned memory on will be the current accelerator.