Performance measurement TensorFlow

Hi,

i have some trained Models on TF2 and i want to measure the performance while executing the inference. I have seen that there is something like that for TensorFlow Lite, where you can measure

Initialization time
Inference time of warmup state
Inference time of steady state
Memory usage during initialization time
Overall memory usage

Is there something for normal TensorFlow?
I use the Python API on Jetson Platform (Linux aarch64)

Thank you very much

Have you seen TensorFlow Profiler: Profil modeli performansı  |  TensorBoard?

1 Like

Check also benchmarks/perfzero at master · tensorflow/benchmarks · GitHub

1 Like