mirror of
https://github.com/ikawrakow/ik_llama.cpp.git
synced 2026-04-21 23:19:22 +00:00
bitnet: scale is per row, not per tensor
This commit is contained in:
@@ -1347,12 +1347,12 @@ static void mul_mat_iq1bn_q8_K64(int n, const void * vx, size_t bx, const DataIn
|
||||
typedef union { float f; uint32_t i; } scale_t;
|
||||
|
||||
scale_t scale;
|
||||
uint16_t u = x[0].extra & 0xff;
|
||||
scale.i = ((((u >> 4) | 0xf0) - 132) << 23) | ((u & 0x0f) << 19);
|
||||
|
||||
for (int ix = 0; ix < nrc_x; ++ix) {
|
||||
|
||||
x = (const block_iq1_bn *)((const char *)vx + ix*bx);
|
||||
uint16_t u = x[0].extra & 0xff;
|
||||
scale.i = ((((u >> 4) | 0xf0) - 132) << 23) | ((u & 0x0f) << 19);
|
||||
|
||||
for (int iy = 0; iy < nrc_y; ++iy) accd[iy] = _mm256_setzero_ps();
|
||||
|
||||
|
||||
Reference in New Issue
Block a user