mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-12-27 20:04:35 +00:00
166 lines
7.1 KiB
C
166 lines
7.1 KiB
C
#pragma once
|
|
|
|
#include "ggml.h"
|
|
|
|
#include <stdint.h>
|
|
#include <assert.h>
|
|
#include <stddef.h>
|
|
|
|
// Super-block size
|
|
#ifdef GGML_QKK_64
|
|
#define QK_K 64
|
|
#define K_SCALE_SIZE 4
|
|
#else
|
|
#define QK_K 256
|
|
#define K_SCALE_SIZE 12
|
|
#endif
|
|
|
|
#ifndef static_assert
|
|
#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201100L)
|
|
#define static_assert(cond, msg) _Static_assert(cond, msg)
|
|
#else
|
|
#define static_assert(cond, msg) struct global_scope_noop_trick
|
|
#endif
|
|
#endif
|
|
|
|
//
|
|
// Super-block quantization structures
|
|
//
|
|
|
|
// 2-bit quantization
|
|
// weight is represented as x = a * q + b
|
|
// 16 blocks of 16 elements each
|
|
// Effectively 2.5625 bits per weight
|
|
typedef struct {
|
|
uint8_t scales[QK_K/16]; // scales and mins, quantized with 4 bits
|
|
uint8_t qs[QK_K/4]; // quants
|
|
ggml_fp16_t d; // super-block scale for quantized scales
|
|
ggml_fp16_t dmin; // super-block scale for quantized mins
|
|
} block_q2_K;
|
|
static_assert(sizeof(block_q2_K) == 2*sizeof(ggml_fp16_t) + QK_K/16 + QK_K/4, "wrong q2_K block size/padding");
|
|
|
|
// 3-bit quantization
|
|
// weight is represented as x = a * q
|
|
// 16 blocks of 16 elements each
|
|
// Effectively 3.4375 bits per weight
|
|
#ifdef GGML_QKK_64
|
|
typedef struct {
|
|
uint8_t hmask[QK_K/8]; // quants - high bit
|
|
uint8_t qs[QK_K/4]; // quants - low 2 bits
|
|
uint8_t scales[2];
|
|
ggml_fp16_t d; // super-block scale
|
|
} block_q3_K;
|
|
static_assert(sizeof(block_q3_K) == sizeof(ggml_fp16_t) + QK_K / 4 + QK_K / 8 + 2, "wrong q3_K block size/padding");
|
|
#else
|
|
typedef struct {
|
|
uint8_t hmask[QK_K/8]; // quants - high bit
|
|
uint8_t qs[QK_K/4]; // quants - low 2 bits
|
|
uint8_t scales[12]; // scales, quantized with 6 bits
|
|
ggml_fp16_t d; // super-block scale
|
|
} block_q3_K;
|
|
static_assert(sizeof(block_q3_K) == sizeof(ggml_fp16_t) + QK_K / 4 + QK_K / 8 + 12, "wrong q3_K block size/padding");
|
|
#endif
|
|
|
|
// 4-bit quantization
|
|
// 8 blocks of 32 elements each
|
|
// weight is represented as x = a * q + b
|
|
// Effectively 4.5 bits per weight
|
|
#ifdef GGML_QKK_64
|
|
typedef struct {
|
|
ggml_fp16_t d[2]; // super-block scales/mins
|
|
uint8_t scales[2]; // 4-bit block scales/mins
|
|
uint8_t qs[QK_K/2]; // 4--bit quants
|
|
} block_q4_K;
|
|
static_assert(sizeof(block_q4_K) == 2*sizeof(ggml_fp16_t) + QK_K/2 + 2, "wrong q4_K block size/padding");
|
|
#else
|
|
typedef struct {
|
|
ggml_fp16_t d; // super-block scale for quantized scales
|
|
ggml_fp16_t dmin; // super-block scale for quantized mins
|
|
uint8_t scales[K_SCALE_SIZE]; // scales and mins, quantized with 6 bits
|
|
uint8_t qs[QK_K/2]; // 4--bit quants
|
|
} block_q4_K;
|
|
static_assert(sizeof(block_q4_K) == 2*sizeof(ggml_fp16_t) + K_SCALE_SIZE + QK_K/2, "wrong q4_K block size/padding");
|
|
#endif
|
|
|
|
// 5-bit quantization
|
|
// 8 blocks of 32 elements each
|
|
// weight is represented as x = a * q + b
|
|
// Effectively 5.5 bits per weight
|
|
#ifdef GGML_QKK_64
|
|
typedef struct {
|
|
ggml_fp16_t d; // super-block scale
|
|
int8_t scales[QK_K/16]; // 8-bit block scales
|
|
uint8_t qh[QK_K/8]; // quants, high bit
|
|
uint8_t qs[QK_K/2]; // quants, low 4 bits
|
|
} block_q5_K;
|
|
static_assert(sizeof(block_q5_K) == sizeof(ggml_fp16_t) + QK_K/2 + QK_K/8 + QK_K/16, "wrong q5_K block size/padding");
|
|
#else
|
|
typedef struct {
|
|
ggml_fp16_t d; // super-block scale for quantized scales
|
|
ggml_fp16_t dmin; // super-block scale for quantized mins
|
|
uint8_t scales[K_SCALE_SIZE]; // scales and mins, quantized with 6 bits
|
|
uint8_t qh[QK_K/8]; // quants, high bit
|
|
uint8_t qs[QK_K/2]; // quants, low 4 bits
|
|
} block_q5_K;
|
|
static_assert(sizeof(block_q5_K) == 2*sizeof(ggml_fp16_t) + K_SCALE_SIZE + QK_K/2 + QK_K/8, "wrong q5_K block size/padding");
|
|
#endif
|
|
|
|
// 6-bit quantization
|
|
// weight is represented as x = a * q
|
|
// 16 blocks of 16 elements each
|
|
// Effectively 6.5625 bits per weight
|
|
typedef struct {
|
|
uint8_t ql[QK_K/2]; // quants, lower 4 bits
|
|
uint8_t qh[QK_K/4]; // quants, upper 2 bits
|
|
int8_t scales[QK_K/16]; // scales, quantized with 8 bits
|
|
ggml_fp16_t d; // super-block scale
|
|
} block_q6_K;
|
|
static_assert(sizeof(block_q6_K) == sizeof(ggml_fp16_t) + QK_K / 16 + 3*QK_K/4, "wrong q6_K block size/padding");
|
|
|
|
// This is only used for intermediate quantization and dot products
|
|
typedef struct {
|
|
float d; // delta
|
|
int8_t qs[QK_K]; // quants
|
|
int16_t bsums[QK_K/16]; // sum of quants in groups of 16
|
|
} block_q8_K;
|
|
static_assert(sizeof(block_q8_K) == sizeof(float) + QK_K + QK_K/16*sizeof(int16_t), "wrong q8_K block size/padding");
|
|
|
|
|
|
// Quantization
|
|
void quantize_row_q2_K_reference(const float * restrict x, block_q2_K * restrict y, int k);
|
|
void quantize_row_q3_K_reference(const float * restrict x, block_q3_K * restrict y, int k);
|
|
void quantize_row_q4_K_reference(const float * restrict x, block_q4_K * restrict y, int k);
|
|
void quantize_row_q5_K_reference(const float * restrict x, block_q5_K * restrict y, int k);
|
|
void quantize_row_q6_K_reference(const float * restrict x, block_q6_K * restrict y, int k);
|
|
void quantize_row_q8_K_reference(const float * restrict x, block_q8_K * restrict y, int k);
|
|
|
|
void quantize_row_q2_K(const float * restrict x, void * restrict y, int k);
|
|
void quantize_row_q3_K(const float * restrict x, void * restrict y, int k);
|
|
void quantize_row_q4_K(const float * restrict x, void * restrict y, int k);
|
|
void quantize_row_q5_K(const float * restrict x, void * restrict y, int k);
|
|
void quantize_row_q6_K(const float * restrict x, void * restrict y, int k);
|
|
void quantize_row_q8_K(const float * restrict x, void * restrict y, int k);
|
|
|
|
// Dequantization
|
|
void dequantize_row_q2_K(const block_q2_K * restrict x, float * restrict y, int k);
|
|
void dequantize_row_q3_K(const block_q3_K * restrict x, float * restrict y, int k);
|
|
void dequantize_row_q4_K(const block_q4_K * restrict x, float * restrict y, int k);
|
|
void dequantize_row_q5_K(const block_q5_K * restrict x, float * restrict y, int k);
|
|
void dequantize_row_q6_K(const block_q6_K * restrict x, float * restrict y, int k);
|
|
void dequantize_row_q8_K(const block_q8_K * restrict x, float * restrict y, int k);
|
|
|
|
// Dot product
|
|
void ggml_vec_dot_q2_K_q8_K(int n, float * restrict s, const void * restrict vx, const void * restrict vy);
|
|
void ggml_vec_dot_q3_K_q8_K(int n, float * restrict s, const void * restrict vx, const void * restrict vy);
|
|
void ggml_vec_dot_q4_K_q8_K(int n, float * restrict s, const void * restrict vx, const void * restrict vy);
|
|
void ggml_vec_dot_q5_K_q8_K(int n, float * restrict s, const void * restrict vx, const void * restrict vy);
|
|
void ggml_vec_dot_q6_K_q8_K(int n, float * restrict s, const void * restrict vx, const void * restrict vy);
|
|
|
|
// Quantization with histogram collection
|
|
size_t ggml_quantize_q2_K(const float * src, void * dst, int n, int k, int64_t * hist);
|
|
size_t ggml_quantize_q3_K(const float * src, void * dst, int n, int k, int64_t * hist);
|
|
size_t ggml_quantize_q4_K(const float * src, void * dst, int n, int k, int64_t * hist);
|
|
size_t ggml_quantize_q5_K(const float * src, void * dst, int n, int k, int64_t * hist);
|
|
size_t ggml_quantize_q6_K(const float * src, void * dst, int n, int k, int64_t * hist);
|
|
|