gpu/cpu leakage

by felfri - opened 4 days ago

4 days ago

there seems to be a computation/memory leakage when running the model. a) casting pipe.to('cuda:2') casts only part of the model to the device and b) when running torch.compile(pipe.transformer) I get "skipping cudagraphs due to skipping cudagraphs due to cpu device."

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

Your need to confirm your account before you can post a new comment.

· Sign up or log in to comment