QuantizationA technique that represents model values with lower precision to reduce memory and inference cost.