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

Vulkan Mixture of Experts (MoE) support (#7628)

Browse source 
* Finish Vulkan mul_mat_id implementation

* Add Vulkan sum_rows and div ops

* Fix MUL_MAT_ID matrix matrix shader

* Fix MUL_MAT_ID matrix vector shader dispatch size

* Fix MUL_MAT_ID matrix vector shader and dispatch code

* Update Vulkan CPU offload for MUL_MAT_ID

* Fix crash when using split mode none and setting a main GPU
This commit is contained in:
0cc4m 2024-06-03 10:59:14 +02:00 • committed by GitHub
parent a10cda58d3
commit 3d7ebf6312
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 73389 additions and 13839 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama.cpp
View file

@ -16372,7 +16372,7 @@ struct llama_context * llama_new_context_with_model(
return nullptr;
}
if (model->split_mode == LLAMA_SPLIT_MODE_NONE) {
ggml_backend_t backend = ggml_backend_vk_init(0);
ggml_backend_t backend = ggml_backend_vk_init(model->main_gpu);
if (backend == nullptr) {
LLAMA_LOG_ERROR("%s: failed to initialize Vulkan backend\n", __func__);
llama_free(ctx);