Are there some differences between .cuda() and to(torch.device('cuda'))?

Eric_K · August 3, 2018, 1:16pm

As mentioned above, we use .cuda() in pytorch 0.3, but we can use .to(torch.device(‘cuda’)) in 0.4 too.
Are there some differences between them? Like speed? Thanks.

albanD · August 3, 2018, 1:19pm

Hi,

No they are the same. .to() is the new version that is more versatile as it take amy device as input.