TensorFlow Lite Micro
A C++ interpreter that runs .tflite models on microcontrollers in one fixed block of RAM.
TensorFlow Lite Micro reads a .tflite FlatBuffer in place, resolves its operators through an op resolver and runs them on tensors planned inside a caller-supplied tensor arena, with no dynamic allocation after setup.
Examples
The
.tflitegesture model is compiled into the firmware as a byte array, and TensorFlow Lite Micro runs it each time a new window of readings arrives.TFLM runs a fixed model: nothing on the board changes the weights.