vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

with forced wmma for cu11

Browse source
This commit is contained in:
Concedo 2025-08-22 17:36:29 +08:00
parent 257992d6b8
commit f545f4df75
1 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml/src/ggml-cuda/fattn.cu
View file

@ -396,9 +396,9 @@ static best_fattn_kernel ggml_cuda_get_best_fattn_kernel(const int device, const
}
//kcpp: use wmma to fix cu11 incoherence
// if (fp16_mma_available(cc) && (ggml_cuda_highest_compiled_arch(cc) <= GGML_CUDA_CC_TURING || cc == GGML_CUDA_CC_TURING)) {
// return BEST_FATTN_KERNEL_WMMA_F16;
// }
if (fp16_mma_available(cc) && (ggml_cuda_highest_compiled_arch(cc) <= GGML_CUDA_CC_TURING || cc == GGML_CUDA_CC_TURING)) {
return BEST_FATTN_KERNEL_WMMA_F16;
}
return BEST_FATTN_KERNEL_MMA_F16;
}