tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-06 17:13:34 -04:00
Code Issues Packages Projects Releases Wiki Activity

ggml : build backends as libraries (#10256)

Browse Source 
* ggml : build backends as libraries

---------

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: R0CKSTAR <xiaodong.ye@mthreads.com>
This commit is contained in:
Diego Devesa
2024-11-14 18:04:35 +01:00
committed by GitHub
parent 4a8ccb37ad
commit ae8de6d50a
191 changed files with 17541 additions and 16879 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cuda/fattn-vec-f16.cuh
View File

@@ -2,9 +2,9 @@
#include "fattn-common.cuh"
template<int D, int ncols, int parallel_blocks, ggml_type type_K, ggml_type type_V, bool use_logit_softcap> // D == head size
#if !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__))
#if !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__))
__launch_bounds__(D, 1)
#endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__))
#endif // !(defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__))
static __global__ void flash_attn_vec_ext_f16(
const char * __restrict__ Q,
const char * __restrict__ K,