I want to measure the inference time of TensorFlow Lite implemented on a Microcontroller (Nano Sense 33). I am beginner to TFLite and would be thankful if anyone can suggest me about the followig. I don’t know about the structure of the TensorFlow Lite C++ Library. I just trained a sine-value predictor model using a Google Colab and then used a Hellow_World example from Arduino to run that example on Nano 33. The trained model is imported into the Arduino using C Array. We have implemented a DWT time measurement class which produces a time elapsed between different parts of the code/program during exexution. But I am not very familiar with the TensorFlow Lite architecture and need help about:
- How can we determine/calculate the inference time of the TFlite? I mean where to initiate the timestamping in the C++ library of the TensorFlow Lite?
- Is it possible to know how much layers are being used inside TFlite? and How to measure that what is the time consumed by each layer?
Thanks in advance! Please guide!