minor tweak for sd

koboldcpp.py

@@ -1786,6 +1786,7 @@ def sd_load_model(model_filename,vae_filename,lora_filename,t5xxl_filename,clip1
     inputs.photomaker_filename = photomaker_filename.encode("UTF-8")
     inputs.img_hard_limit = args.sdclamped
     inputs.img_soft_limit = args.sdclampedsoft
+    inputs.lora_apply_mode = 0 #auto for now
     inputs = set_backend_props(inputs)
     ret = handle.sd_load_model(inputs)
     return ret

otherarch/sdcpp/lora.hpp

@@ -707,7 +707,7 @@ struct LoraModel : public GGMLRunner {
             if (applied_lora_tensors.find(kv.first) == applied_lora_tensors.end()) {
                 if (!at_runntime) {
                     LOG_WARN("unused lora tensor |%s|", kv.first.c_str());
-                    print_ggml_tensor(kv.second, true);
+                    //print_ggml_tensor(kv.second, true);
                 }
             } else {
                 applied_lora_tensors_count++;