Mbed Torch Fusion OS

MbedTorch Fusion OS: Seamlessly merges Mbed OS with Torch ML models for the P-Nucleo-WB55RG on Arm Cortex M4. Empowering edge computing with efficient ML deployment.

Tutorial: https://youtu.be/jc_GCxAAO20

1. Build ExecuTorch libraries

git clone https://github.com/ChristophKarlHeck/mbed-torch-fusion-os.git

cd mbed-torch-fusion-os/utils/scripts

./build_et_libs.sh

cd ../../

2. Install pyOCD in mbed-os venv

source mbed-os/venv/bin/activate

python3 -m pip install -U pyocd

pyocd pack --update

pyocd pack --install stm32wb55rg

deactivate

3. Convert softmax.pte file to model_pte.h

cd executorch

source .executorch/bin/activate

export PATH="$(pwd)/third-party/flatbuffers/cmake-out:${PATH}"

"$(pwd)/build/install_flatc.sh"

export PATH="$(pwd)/../gcc-arm-none-eabi-10.3-2021.10/bin:${PATH}"

hash arm-none-eabi-gcc

cd ..

python3 executorch/examples/arm/executor_runner/pte_to_header.py --pte executorch/softmax.pte --outdir .

4. Connect with serial port

Open a new terminal. Plug in Nucleo-WB55RG. Install Mbed CLI 2 and use it to find serial port.

sudo apt-get install python3-dev

python3 -m pip install mbed-tools

mbed-tools detect

Remember serial port. Use minicom to see console output.

minicom -D /dev/ttyACM0

ctrl a & x to exit

5. Build and flash using VS Code

https://github.com/mbed-ce/mbed-os/wiki/Project-Setup:-VS-Code