Expand description
Quantization conversion functions
This module contains functions for converting between floating-point and quantized data, including matrix and vector quantization/dequantization and fake quantization.
Functionsยง
- dequantize_
matrix - Dequantize a matrix back to floating-point
- dequantize_
vector_ public - Dequantize a vector back to floating-point (public API version)
- fake_
quantize - Apply fake quantization to a floating-point matrix
- fake_
quantize_ vector - Apply fake quantization to a floating-point vector
- quantize_
matrix - Quantize a floating-point matrix to a lower precision representation
- quantize_
matrix_ per_ channel - Quantize a floating-point matrix using per-channel quantization
- quantize_
vector - Quantize a floating-point vector to a lower precision representation