quantized_vector.hpp File Reference

INT8/INT4 quantized vector storage for AI embeddings with SIMD acceleration. More...

#include "signet/types.hpp"
#include "signet/error.hpp"
#include <algorithm>
#include <atomic>
#include <cmath>
#include <cstdint>
#include <cstdio>
#include <cstring>
#include <limits>
#include <string>
#include <vector>

Go to the source code of this file.

Classes
struct	signet::forge::QuantizationParams
	Parameters that fully describe a quantization mapping. More...
class	signet::forge::Quantizer
	Quantizes float32 vectors to INT8 or INT4 representation. More...
class	signet::forge::Dequantizer
	Dequantizes INT8/INT4 quantized vectors back to float32. More...
class	signet::forge::QuantizedVectorWriter
	Accumulates float32 vectors, quantizes them, and produces FIXED_LEN_BYTE_ARRAY page data suitable for Parquet column chunks. More...
class	signet::forge::QuantizedVectorReader
	Reads quantized page data (FIXED_LEN_BYTE_ARRAY) and dequantizes to float32 on demand. More...
struct	signet::forge::QuantizedVectorReader::RawResult
	Result of a raw (non-dequantized) page read. More...

Namespaces
namespace	signet
namespace	signet::forge

Enumerations
enum class	signet::forge::QuantizationScheme : int32_t { signet::forge::SYMMETRIC_INT8 = 0 , signet::forge::ASYMMETRIC_INT8 = 1 , signet::forge::SYMMETRIC_INT4 = 2 }
	Identifies the quantization method used for vector compression. More...

Detailed Description

INT8/INT4 quantized vector storage for AI embeddings with SIMD acceleration.

Definition in file quantized_vector.hpp.