vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fixup! fixup! fixup! CUDA: add FP32 FlashAttention vector kernel

Browse source
This commit is contained in:
Johannes Gäßler 2024-05-11 15:34:09 +02:00
parent f3c3eafa6e
commit aa9cbd7660
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-cuda/fattn-vec-f16.cu
View file

@ -342,7 +342,7 @@ void ggml_cuda_flash_attn_ext_vec_f16_no_mma(ggml_backend_cuda_context & ctx, gg
ggml_tensor * KQV = dst; ggml_tensor * KQV = dst;
const int32_t precision = KQV->op_params[1]; const int32_t precision = KQV->op_params[2];
GGML_ASSERT(precision == GGML_PREC_DEFAULT); GGML_ASSERT(precision == GGML_PREC_DEFAULT);
GGML_ASSERT(Q->ne[0] == 64 || Q->ne[0] == 128 && "FlashAttention without tensor cores only supports head sizes 64 and 128."); GGML_ASSERT(Q->ne[0] == 64 || Q->ne[0] == 128 && "FlashAttention without tensor cores only supports head sizes 64 and 128.");