Quantization-Aware Training (QAT) is a method that models low-precision arithmetic (like INT4 or INT8) during training or fine-tuning - View it on GitHub