Module quantization 

Quantization Support for Model Compression

This module provides quantization techniques to compress knowledge graph embeddings by reducing precision from float32 to int8/int4, significantly reducing model size and improving inference speed.

Structs

ModelQuantizer

Model quantizer

QuantizationConfig

Quantization configuration

QuantizationParams

Quantization parameters

QuantizationStats

Quantization statistics

QuantizedTensor

Quantized tensor representation

Enums

BitWidth

Quantization bit width

QuantizationScheme

Quantization scheme