TFLM Template

This project builds TensorFlow Lite Micro libraries and a simple runner for your models.

Setup

Install dependencies before building:

python -m venv venv
. venv/bin/activate
pip install -r requirements.txt

Build

Run make to generate static libraries for both the host (Linux) and Cortex-M7 (with CMSIS-NN):

make

The build pulls in TFLM sources and produces the libraries under gen/, including:

• gen/linux_x86_64_debug_gcc/lib/libtensorflow-microlite.a
• gen/f7/lib/libtensorflow-microlite.a
• gen/h7/lib/libtensorflow-microlite.a

To add your own models, drop .tflite files into src/models. The generator will include them on the next make and expose per-model entrypoints named tflm_main_<model_name>, each taking (uint8_t* tensor_arena, int tensor_arena_size, uint32_t (*get_time_ms)()) so you can supply your own arena buffer and an millisecond time getter (e.g., HAL_GetTick on STM32).

Run models on host

After make, execute the host runner to invoke all configured models:

./tflm_main