Quantization

A technique for compressing AI models that involves reducing the precision of parameters (for example, going from 32 bits to 8 bits) to decrease model size and speed up inference, often used in embedded applications.

‍