Open chin-jey opened 1 year ago
The underlying operations should be the same (as they are handled by libtorch both on this version and on the PyTorch version) so I would expect performance to be reasonably in line.
what is advantage of impementing this in rust if there any. Thanks
Hi @LaurentMazare, have you benchmarked this against huggingface's python diffusers? Should I expected it to be any faster? If yes, can you give me some intuition on the reason?