Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-10-30 06:30:15 +01:00
Code Issues Packages Projects Releases Wiki Activity

ggml : fix q2_k bpw in comments (ggml/680)

Browse Source
This commit is contained in:
Georgi Gerganov 2024-01-05 15:36:04 +02:00
parent 1bf681f90e
commit d061bf9405
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml-quants.h
View File

@ -70,7 +70,7 @@ static_assert(sizeof(block_q8_1) == 2*sizeof(float) + QK8_1, "wrong q8_1 block s
// 2-bit quantization // 2-bit quantization
// weight is represented as x = a * q + b // weight is represented as x = a * q + b
// 16 blocks of 16 elements each // 16 blocks of 16 elements each
// Effectively 2.5625 bits per weight // Effectively 2.625 bits per weight
typedef struct { typedef struct {
uint8_t scales[QK_K/16]; // scales and mins, quantized with 4 bits uint8_t scales[QK_K/16]; // scales and mins, quantized with 4 bits
uint8_t qs[QK_K/4]; // quants uint8_t qs[QK_K/4]; // quants