MayaFlux 1.0.0
Digital-First Multimedia Processing Framework
Loading...
Searching...
No Matches
MayaFlux::Kakshya::TensorQuantization Struct Reference

Affine quantization: real = scale * (stored - zero_point). More...

#include <TensorSpec.hpp>

Collaboration diagram for MayaFlux::Kakshya::TensorQuantization:

Public Attributes

std::optional< size_t > axis
std::vector< double > scale
std::vector< int64_t > zero_point

Detailed Description

Affine quantization: real = scale * (stored - zero_point).

Per tensor when axis is empty (one scale), otherwise per slice along axis (one scale per index of that axis). zero_point is empty for symmetric quantization or matches scale in length.

Definition at line 70 of file TensorSpec.hpp.


The documentation for this struct was generated from the following file: