Skip to main content

quantize_model

Function quantize_model 

Source
pub fn quantize_model(model: &Model, bits: u32) -> Model
Expand description

Simple weight quantization: clamp weights to int8 range then dequantize. This simulates the effect of lower-precision storage.