Hi @Raymond, Raspberry Pi is running the latest available for it TFLite version (v2.5.0) which is older than the recent TF builds.
If you are getting this error, please try using TensorFlow 2.5-2.7 for converting your model to TFLite format.
The Burnout Score is computed as a geometric mean of the Compute, Throttling and Power Efficiency scores.
Compute, Throttling and Power Efficiency scores are proportional to the corresponding FPS and FPS/Watt values.
You should generally send us the product you've developed, and we will then benchmark it in our lab and publish the scores. The publication of the results is free of charge for everyone. For more information, please contact us by email (you can find it on the website).