ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-30 19:31:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix it

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-09-02 14:35:05 +03:00
parent be2694eb68
commit 32e223df46
1 changed files with 0 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/src/ggml-cuda/fattn-mma-f16.cuh
View File

@@ -1035,9 +1035,6 @@ static __global__ void flash_attn_mma_ext_f16(
int kb0_start_kernel = kb0_start * kb_niter; int kb0_start_kernel = kb0_start * kb_niter;
int kb0_stop_kernel = kb0_stop * kb_niter; int kb0_stop_kernel = kb0_stop * kb_niter;
if (bounds) { if (bounds) {
if (kb0_start_kernel*KQ_per_iter >= bounds[jt].y || kb0_stop_kernel*KQ_per_iter < bounds[jt].x) {
return;
}
kb0_start_kernel = max(kb0_start_kernel, bounds[jt].x / KQ_per_iter); kb0_start_kernel = max(kb0_start_kernel, bounds[jt].x / KQ_per_iter);
kb0_stop_kernel = min(kb0_stop_kernel, bounds[jt].y / KQ_per_iter); kb0_stop_kernel = min(kb0_stop_kernel, bounds[jt].y / KQ_per_iter);
} }