sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-03-17 16:44:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: fix flash attention dot product precision (#20589)

Browse Source
This commit is contained in:
Ruben Ortlam
2026-03-16 10:45:49 +01:00
committed by GitHub
parent de8f01c2d7
commit 46dba9fce8
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp
View File

@@ -245,7 +245,7 @@ void main() {
#endif
}
[[unroll]] for (uint32_t r = 0; r < rows_per_thread; ++r) {
Sf[r][c] += ACC_TYPE(dot(Q_cache[r], K_Tf));
Sf[r][c] += dot(ACC_TYPEV4(Q_cache[r]), ACC_TYPEV4(K_Tf));
}
}
}
@@ -270,7 +270,7 @@ void main() {
#endif
}
[[unroll]] for (uint32_t r = 0; r < rows_per_thread; ++r) {
Sf[r][c] += ACC_TYPE(dot(Qf[tile_row(r) * qf_stride + d * D_split + d_tid], K_Tf));
Sf[r][c] += dot(ACC_TYPEV4(Qf[tile_row(r) * qf_stride + d * D_split + d_tid]), ACC_TYPEV4(K_Tf));
}
}
}