Module adaptive

Module adaptive 

Source
Expand description

Adaptive quantization module

This module provides functionality to automatically select the best quantization scheme based on high-level requirements (accuracy vs compression).

§Performance

Adaptive quantization introduces negligible overhead (zero cost in most cases) compared to direct scheme usage, as the selection logic is effectively inlined by the compiler.

Enums§

AccuracyTarget
Target accuracy levels for adaptive quantization
CompressionTarget
Target compression levels for adaptive quantization

Functions§

quantize_adaptive
Quantize an embedding based on an accuracy target
quantize_with_target
Quantize an embedding based on a compression target