keep printing out of kcpp_backend

Author: wbruna

gpttype_adapter.cpp

@@ -3215,7 +3215,11 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
             printf("Main GPU device: Try set to %d\n",kcpp_parseinfo_maindevice);
         }
 
-        kcpp_backend_cuda_set_mul_mat_q(inputs.use_mmq);
+        if(kcpp_backend_check(KCPP_BACKENDS_USE_CUDA)) {
+            printf("CUDA MMQ: %s\n",(inputs.use_mmq?"True":"False"));
+            printf("---\nInitializing CUDA/HIP, please wait, the following step may take a few minutes (only for first launch)...\n---\n");
+            kcpp_backend_cuda_set_mul_mat_q(inputs.use_mmq);
+        }
 
         model_params.main_gpu = kcpp_parseinfo_maindevice;
         model_params.split_mode = (inputs.splitmode>0?((llama_split_mode)(inputs.splitmode)):llama_split_mode::LLAMA_SPLIT_MODE_LAYER);

kcpp_backend.cpp

@@ -141,7 +141,6 @@ void kcpp_backend_cuda_ggmlv2_transform_tensor(ggml_v2_tensor * tensor)
 void kcpp_backend_cuda_ggmlv3_set_main_device(int device)
 {
     #if defined(GGML_USE_CUDA)
-    printf("CUBLAS v3: Set main device to %d\n",device);
     ggml_v3_cuda_set_main_device(device);
     #endif
 }
@@ -212,8 +211,6 @@ void kcpp_backend_cuda_ggmlv3_set_scratch_size(size_t scratch_size)
 void kcpp_backend_cuda_set_mul_mat_q(int use_mmq)
 {
     #if defined(GGML_USE_CUDA)
-    printf("CUDA MMQ: %s\n",(use_mmq?"True":"False"));
-    printf("---\nInitializing CUDA/HIP, please wait, the following step may take a few minutes (only for first launch)...\n---\n");
     ggml_cuda_set_mul_mat_q(use_mmq);
     #endif
 }
@@ -225,7 +222,3 @@ void kcpp_backend_hip_initialize()
     #endif
 }
 
-
-
-
-

otherarch/llama_v3.cpp

@@ -1162,6 +1162,7 @@ static void llama_v3_model_load_internal(
 
     if (kcpp_backend_check(KCPP_BACKENDS_USE_CUDA)) {
     LLAMA_V3_LOG_INFO("%s: using CUDA for GPU acceleration\n", __func__);
+    printf("CUBLAS v3: Set main device to %d\n",main_gpu);
     kcpp_backend_cuda_ggmlv3_set_main_device(main_gpu);
     kcpp_backend_cuda_ggmlv3_set_mul_mat_q(mul_mat_q);
     LLAMA_V3_BACKEND_OFFLOAD = GGML_V3_BACKEND_GPU;