Pipe.to("cuda") not working

I found there’s an update pipe for parallel computing using GPU, use this to improve it,

Original code:
pipe.to(“gpu”)

Then replace with:
pipe.to(device=“cuda”, dtype=torch.float16)