Open jignesh222 opened 1 month ago
I think the hardest part is to find a wasi-nn backend that works with ZephyrOS. Right now, WAMR has three backends: tensorflow-lite, openvino, and llama.cpp. You'll have to make one of these work on ZephyrOS. I'm not sure if any of them have been adapted for it yet.
Afterwards, you just need to compile WAMR's source code again with wasi-nn support turned on (using WAMR_BUILD_WASI_NN
, WAMR_BUILD_WASI_EPHEMERAL_NN
and the backend you picked), sort out any issues with the compilation, like warnings or errors, and then test it out.
I'm expecting WASI-NN based examples for Zephyr Platform to run small NN models in controller environment using WAMR.
Can I get any solution on this?