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

prevent lora with clblast

Browse source
This commit is contained in:
Concedo 2023-10-25 15:18:03 +08:00
parent 30d1017021
commit c9983a72d6
2 changed files with 6 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

5
llama.cpp
View file

@ -8854,11 +8854,12 @@ static int llama_apply_lora_from_file_internal(
offload_func_t offload_func = llama_nop;
offload_func_t offload_func_force_inplace = llama_nop;
#ifdef GGML_USE_CUBLAS
#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST)
if (dest_t->backend == GGML_BACKEND_GPU || dest_t->backend == GGML_BACKEND_GPU_SPLIT) {
if (dest_t->type != GGML_TYPE_F16) {
printf("\nError: the simultaneous use of LoRAs and GPU acceleration is only supported for f16 models\n");
throw std::runtime_error(format(
"%s: error: the simultaneous use of LoRAs and GPU acceleration is only supported for f16 models", __func__));
"%s: error: lora failed", __func__));
}
offload_func = ggml_cuda_assign_buffers;
offload_func_force_inplace = ggml_cuda_assign_buffers_force_inplace;

5
otherarch/llama_v3.cpp
View file

@ -3763,11 +3763,12 @@ int llama_v3_apply_lora_from_file_internal(const struct llama_v3_model & model,
offload_func_t offload_func = llama_v3_nop;
offload_func_t offload_func_force_inplace = llama_v3_nop;
#ifdef GGML_USE_CUBLAS
#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST)
if (dest_t->backend == GGML_BACKEND_GPU || dest_t->backend == GGML_BACKEND_GPU_SPLIT) {
if (dest_t->type != GGML_TYPE_F16) {
printf("\nError: the simultaneous use of LoRAs and GPU acceleration is only supported for f16 models\n");
throw std::runtime_error(format_old(
"%s: error: the simultaneous use of LoRAs and GPU acceleration is only supported for f16 models", __func__));
"%s: error: lora failed", __func__));
}
offload_func = ggml_cuda_assign_buffers;
offload_func_force_inplace = ggml_cuda_assign_buffers_force_inplace;