#quantization-and-numeric-precision

Model Quantization

Quantization is a model compression technique that reduces numerical precision of weights and activations from floating-point to lower-bit representations, decreasing model size and computational cost