An automated method for optimizing quantized models by adjusting weight rounding to minimize accuracy loss during compression.