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

Merge remote-tracking branch 'llama.cpp/try-fix-3869' into concedo_experimental

Browse source
This commit is contained in:
Concedo 2023-11-01 18:29:08 +08:00
commit bcb397953f
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

@ -7319,7 +7319,7 @@ static void ggml_cuda_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1
} else if (all_on_device && !use_tensor_cores && src0->type == GGML_TYPE_F16 && !ggml_is_contiguous(src0) && !ggml_is_transposed(src1) && src1->ne[1] == 1) {
// KQV single-batch
ggml_cuda_mul_mat_vec_nc(src0, src1, dst);
} else if (all_on_device && src0->type == GGML_TYPE_F16 && src1->type == GGML_TYPE_F32 && !ggml_is_transposed(src0) && !ggml_is_transposed(src1)) {
} else if (all_on_device && use_tensor_cores && src0->type == GGML_TYPE_F16 && src1->type == GGML_TYPE_F32 && !ggml_is_transposed(src0) && !ggml_is_transposed(src1)) {
// KQ + KQV multi-batch
ggml_cuda_mul_mat_mat_batched_cublas(src0, src1, dst);
} else if (src0->type == GGML_TYPE_F32) {