Weights compression using Decompression Unit on STM32N6 NPU

[AMurz.1]

Hi,

Is there a way to enable weights compression while also using the STM32N6 NPU ?

I saw there is Decompression unit in the reference manual where it tells the NPU compiler is the tool that generate codebooks offline (there is also mention of it in lib_aton.h too):

This would help to reduce the model size in RAM.

I've tried with models using a single channel in 1x1 Conv, but maybe this feature can only be used with some specific conditions ?

Thanks!

Alexis Murzeau

[Julian E. · Accepted solution]

Hi @AMurz.1,

The compression is not supported by the ST Edge AI Core CLI using the NPU.

ref: https://stedgeai-dc.st.com/assets/embedded-docs/command_line_interface.html 

For any other MCU you can use this -c option to explicitly use the compression.

For the NPU, the support of compression will most likely come at some point. I know that the first level of developer working on it have some early commands/tools to play with it, but it is not tested and not even available internally.

have a good day,

Julian