What is model quantization in PyTorch?

Quantization reduces model precision to improve speed and reduce size.

PyTorch supports dynamic, static, and QAT quantization, converting FP32 weights into INT8 representations while preserving accuracy through calibration or retraining.