Getting Error "REDUCE_PROD" when trying to run Tensorflow Lite on an ESP32

TensorFlow General Discussion
1

Hey,

im trying to run a Tensorflow Lite Model on an ESP32 to do predictive maintnace.
I tried out the hello world exmple an everything worked fine. Now i changed the hello world model to my model and this Error appeared:
“Didn’t find op for bultin opcode “REDUCE_PROD” version “1”. An older version of this builtin might be supported. Are you unsing an old TFLite binary with a newer model?”
I tried to train my model using Tensorflow 2.4.0 because this is the same version in TFLite. But it throws the same error.
Im using the Arduino 2.0.2 IDE.
Does someone know whats going on here?

3

@Palettenbrett,

Welcome to the Tensorflow Forum!

It seems REDUCE_PROD was not included in Micro. Please refer to Porting Reference Ops from Lite to Micro may help you. Thank you!

5

Thank you for the quick answere,

so i need to port his Op myself? Is there a good tutorial on this somewhere?

6

You can refer to Porting Reference Ops from Lite to Micro : this is a guide to porting reference ops from Lite to Micro. It explains, step-by-step, the recommended code changes and the process for submitting them for review and acceptance.

Thank you!