Aicraft
Skip to main content

Edge Deployment

Deploy Aicraft models on microcontrollers and embedded devices.

Overview​

Aicraft is designed for edge deployment from the ground up. With zero dependencies, a ~150 KB binary, and INT8 quantisation support, it runs on everything from x86 servers to ARM Cortex-M MCUs.

Quantisation Workflow​

#include "aicraft/aicraft.h"

// After training...
// Quantise the model to INT8
ac_quantize_model(net, 2, AC_QUANT_INT8);

// Save quantised model
ac_serialize(net, 2, "model_int8.bin");

Deployment Targets​

TargetCompilerFlags
x86_64 Linux/WinGCC / Clang / MSVC-O3 -mavx2
ARM Cortex-A (RPi)arm-linux-gnueabihf-gcc-O3 -mfpu=neon
ARM Cortex-M (STM32)arm-none-eabi-gcc-O2 -mcpu=cortex-m4
RISC-Vriscv64-linux-gnu-gcc-O3

Memory Budget​

For constrained devices, configure the arena size at init:

ac_init_with_arena(64 * 1024);  // 64 KB arena

Cross-Compilation Example​

# For Raspberry Pi
arm-linux-gnueabihf-gcc -O3 -mfpu=neon \
    inference.c -I./include -o inference

# For STM32
arm-none-eabi-gcc -O2 -mcpu=cortex-m4 -mthumb \
    inference.c -I./include -o inference.elf

Model Size (MNIST, 784→128→10)​

FormatSize
Float32 (original)~400 KB
INT8 (quantised)~100 KB
Binary (code + model)~150 KB