4 ms·8-bit quantization on TensorFlow Lite: faster inference and 4x less memory2 points by austin_kodra 8y ago