ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix ggml_cuda_fattn_is_supported (#968)

Browse Source 
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow
2025-11-16 11:50:29 +02:00
committed by GitHub
parent b40d11b22d
commit 37d72f9878
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cuda/fattn.cu
View File

@@ -169,7 +169,7 @@ bool ggml_cuda_fattn_is_supported(ggml_backend_cuda_context & ctx, const ggml_te
const bool mma_faster_for_bs1 = new_mma_available(cc) && gqa_opt_applies && !(Q->ne[1] == 1 && n_swa > 0);
const bool can_use_vector_kernel = Q->ne[0] <= 256 && Q->ne[0] % (2*WARP_SIZE) == 0;
if (Q->ne[1] == 1 && can_use_vector_kernel && !mma_faster_for_bs1 && !ggml_is_quantized(K->type) && !ggml_is_quantized(V->type)) {
return ggml_cuda_fattn_tile_f32_is_supported(ctx, dst);
return ggml_cuda_fattn_vec_f32_is_supported(ctx, dst);
}
if (new_mma_available(cc) && (Q->ne[0] == 576 || (K->ne[0] == 192 && V->ne[0] == 128))) {