vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Another attempt to fix the Windows builds

Browse source
This commit is contained in:
Iwan Kawrakow 2024-02-23 10:38:02 +02:00
parent 436a146f98
commit 303f3f3258
1 changed files with 10 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

20
ggml-quants.c
View file

@ -11173,20 +11173,20 @@ static void quantize_row_iq3_s_impl(int block_size, const float * restrict x, vo
} }
} }
#define IQ3S_BLOCK_SIZE 32
size_t quantize_iq3_s(const float * src, void * dst, int nrow, int n_per_row, int64_t * hist, const float * quant_weights) { size_t quantize_iq3_s(const float * src, void * dst, int nrow, int n_per_row, int64_t * hist, const float * quant_weights) {
(void)hist; (void)hist;
GGML_ASSERT(n_per_row%QK_K == 0); GGML_ASSERT(n_per_row%QK_K == 0);
int nblock = n_per_row/QK_K; int nblock = n_per_row/QK_K;
const int block_size = 32; float scales[QK_K/IQ3S_BLOCK_SIZE];
float scales[QK_K/block_size]; float weight[IQ3S_BLOCK_SIZE];
float weight[block_size]; float xval[IQ3S_BLOCK_SIZE];
float xval[block_size]; int8_t L[IQ3S_BLOCK_SIZE];
int8_t L[block_size]; int8_t Laux[IQ3S_BLOCK_SIZE];
int8_t Laux[block_size]; float waux[IQ3S_BLOCK_SIZE];
float waux[block_size]; bool is_on_grid[IQ3S_BLOCK_SIZE/4];
bool is_on_grid[block_size/4]; bool is_on_grid_aux[IQ3S_BLOCK_SIZE/4];
bool is_on_grid_aux[block_size/4]; uint8_t block_signs[IQ3S_BLOCK_SIZE/8];
uint8_t block_signs[block_size/8];
char * qrow = (char *)dst; char * qrow = (char *)dst;
for (int row = 0; row < nrow; ++row) { for (int row = 0; row < nrow; ++row) {
quantize_row_iq3_s_impl(32, src, qrow, n_per_row, quant_weights, quantize_row_iq3_s_impl(32, src, qrow, n_per_row, quant_weights,