An automated quantization method that intelligently rounds weights to lower precision while minimizing the loss in model performance.