CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-03 09:22:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

cuda : increase max block size to 1024

Browse Source
This commit is contained in:
Georgi Gerganov
2023-11-30 20:40:47 +02:00
parent 62532c05aa
commit 6b86bcffac
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml-cuda.cu
View File

@@ -443,7 +443,7 @@ static_assert(sizeof(block_q6_K) == sizeof(ggml_fp16_t) + 13*QK_K/16, "wrong q6_
#define CUDA_SCALE_BLOCK_SIZE 256
#define CUDA_CLAMP_BLOCK_SIZE 256
#define CUDA_ROPE_BLOCK_SIZE 256
#define CUDA_SOFT_MAX_BLOCK_SIZE 512
#define CUDA_SOFT_MAX_BLOCK_SIZE 1024
#define CUDA_ALIBI_BLOCK_SIZE 32
#define CUDA_DIAG_MASK_INF_BLOCK_SIZE 32
#define CUDA_QUANTIZE_BLOCK_SIZE 256