Closed fpedd closed 3 years ago
@fabianpedd Thanks for the question. The differences are listed in https://github.com/ARM-software/CMSIS_5/tree/develop/CMSIS/NN#legacy-vs-tfl-micro-compliant-apis . There is more to it than naming convention and the current supported ones are the ones with _s in the name that support the int8/16 spec of TensorFlow Lite for Microcontrollers.
You could have a look at the blog https://community.arm.com/arm-community-blogs/b/ai-and-ml-blog/posts/image-recognition-on-arm-powered-microcontrollers on how to get started..
@felix-johnny thank you.
I am just getting started with CMSIS NN and stumbled upon functions with similar names but different endings, specifically
*_q{7,15}
vs*_s{8,16}
. Both appear to be using theq{7,15,31}
fixed-point data types. What is the difference between theq
ands
versions? Is it just naming convention/legacy, or is there more to it? Tensorflow appears to be using only thes
variant.