ATen icon indicating copy to clipboard operation
ATen copied to clipboard

Published 20 hours ago • verified publisher icon zdevito

Network inference speed

Open pharrellyhy opened this issue 6 years ago • 0 comments

Hi,

I have a neural network which expects input size as (batch_size, 1, 112, 112). If batch size is 1, it takes about 2ms for a forward pass. If I increase batch size to 20, it takes about 35ms which is much slower than the Pytorch counterpart which is ~4ms. Is it expected in ATen?

pharrellyhy avatar Jun 28 '18 07:06 pharrellyhy