It'd be really nice to know where there are performance bottlenecks in your model. I think tensorflow 1.0 added some stuff that would make this relatively easy to diagnose; can we put in some simple tooling that tells you which layers take up the majority of your computation time?
It'd be really nice to know where there are performance bottlenecks in your model. I think tensorflow 1.0 added some stuff that would make this relatively easy to diagnose; can we put in some simple tooling that tells you which layers take up the majority of your computation time?