Inference & performance · Concept

Quantization

Represent model values with reduced precision to lower memory or compute requirements.