Module quantizer 

TensorQuantizer — symmetric and asymmetric quantization for tensor values.

Provides calibration-based quantization with INT8 and INT4 support, including quantize/dequantize roundtrips and MSE error measurement.

Examples

use ipfrs_tensorlogic::quantizer::{TensorQuantizer, QuantMode, QuantBits};

let mut quantizer = TensorQuantizer::new();
quantizer.calibrate(&[1.0, -0.5, 0.3, -1.0]);

let params = quantizer.compute_params(QuantMode::Symmetric, QuantBits::Int8).expect("example: should succeed in docs");
let quantized = TensorQuantizer::quantize(&[0.5, -0.5], &params);
let restored = TensorQuantizer::dequantize(&quantized, &params);

Structs

QuantParams

Parameters computed from calibration data for quantization/dequantization.

QuantizerStats

Statistics about the quantizer’s calibration state.

TensorQuantizer

Calibration-based tensor quantizer.

Enums

QuantBits

Quantization bit width.

QuantMode

Quantization mode.