I have a bit of a strange use case, where I'm trying to run OpenWakeWord on Apple Silicon and tflite-runtime isn't available for it. I am running tensorflow with tensorflow-metal to take advantage of the chip's capabilities, but OWW doesn't attempt to fall back to tensorflow, instead opting for onnx.

It would be nice for full tensorflow to be an option, even if it's just a configurable option with extra dependencies.

Thank you!