wylab/llama.cpp
0
0
Fork 1
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-06 17:57:44 +02:00
Code Issues Packages Projects Releases Wiki Activity

ggml-cpu/ops: vectorize flash-attention V-cache F16 to F32 conversion (#26947)

Browse Source 
Co-authored-by: jinzihao <jinzihao.jzh@alibaba-inc.com>
This commit is contained in:
jinzihao
2026-08-13 12:04:46 +03:00
committed by GitHub
co-authored by jinzihao
parent 154d57af3e
commit eeae28b67e
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cpu/ops.cpp
+1 -1
View File
@@ -8941,7 +8941,7 @@ static void ggml_compute_forward_flash_attn_ext_tiled(
for (int tk = 0; tk < kv_tile; tk++) {
const char * v_data = (const char *)v->data + (ic + tk)*nbv1 + iv2*nbv2 + iv3*nbv3;
if (kv_type == GGML_TYPE_F16) {
ggml_fp16_to_fp32_row((const ggml_fp16_t *)v_data, V32 + tk * DV, DV);
ggml_cpu_fp16_to_fp32((const ggml_fp16_t *)v_data, V32 + tk * DV, DV);
} else {
memcpy(V32 + tk * DV, v_data, DV * sizeof(float));
}