Performance of Fused vs Non Fused convolutions during inference

11 Views Asked by At

My Network is currently running quite slowly on tfjs and, while trying to understand why other larger models (in size) run faster, I saw that other networks use fused layers.

Does anyone knows what would be the performance difference in speed using FusedConvolutions vs non Fused ones in tfjs during inference?

Thanks

0

There are 0 best solutions below