The MLTK supports C++ development and comes with several C++ Python wrappers.
C++ Python wrappers are C++ libraries that have an additional interface which enables them to be invoked from a Python script. This allows for sharing source code between embedded targets and host model training scripts.
The Python wrappers use PyBind11 to manage converting data between Python and C++.
The source code for the wrappers may be found on Github at mltk/cpp
The following wrappers are available:
- Audio Feature Generator - Allows for sharing spectrogram generation algorithms between model training scripts and embedded targets
- Tensorflow-Lite Micro - Allows for running Tensorflow-Lite Micro interpreter from Python
- MVP Kernels - Allows for running MVP hardware accelerated Tensorflow-Lite Micro kernels from Python
.. toctree::
:maxdepth: 1
:hidden:
./audio_feature_generator_wrapper
./tflite_micro_wrapper
./mvp_wrapper