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

ggml : update ggml_soft_max_ext() CUDA, SYCL

Browse source
This commit is contained in:
Georgi Gerganov 2024-05-10 11:12:19 +03:00
parent 7fdca3348c
commit d0592d495d
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
4 changed files with 38 additions and 74 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml-vulkan.cpp
View file

@ -3830,9 +3830,8 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
return nullptr;
case GGML_OP_SOFT_MAX:
GGML_ASSERT(!src1 || src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16);
GGML_ASSERT(!src2 || src2->type == GGML_TYPE_F32 || src2->type == GGML_TYPE_F16);
if (src0->type == GGML_TYPE_F32 && (src1 == nullptr || src1->type == GGML_TYPE_F32) && (src2 == nullptr || src2->type == GGML_TYPE_F32) && dst->type == GGML_TYPE_F32) {
if (src0->type == GGML_TYPE_F32 && (src1 == nullptr || src1->type == GGML_TYPE_F32) && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_soft_max_f32;
}
if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F16 && src2->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F32) {