Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

CUDA: fix Gemma E4B MTP FlashAttention (#25148)

Browse Source 
* CUDA: fix Gemma E4B MTP FlashAttention

* remove unused template declaration
This commit is contained in:
Johannes Gäßler
2026-06-30 14:06:54 +02:00
committed by GitHub
parent f708a5b2ca
commit e495d1e748
8 changed files with 23 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cuda/fattn-mma-f16.cuh
+4
View File
@@ -2003,6 +2003,10 @@ DECL_FATTN_MMA_F16_CASE_ALL_NCOLS2(112, 112, 64)
DECL_FATTN_MMA_F16_CASE_ALL_NCOLS2(128, 128, 64)
DECL_FATTN_MMA_F16_CASE_ALL_NCOLS2(256, 256, 64)
extern DECL_FATTN_MMA_F16_CASE(512, 512, 4, 2);
extern DECL_FATTN_MMA_F16_CASE(512, 512, 8, 2);
extern DECL_FATTN_MMA_F16_CASE(512, 512, 16, 2);
extern DECL_FATTN_MMA_F16_CASE(512, 512, 32, 2);
extern DECL_FATTN_MMA_F16_CASE(512, 512, 2, 4);
extern DECL_FATTN_MMA_F16_CASE(512, 512, 4, 4);
extern DECL_FATTN_MMA_F16_CASE(512, 512, 8, 4);