Update app.py

app.py

@@ -204,36 +204,43 @@ def run_lora(prompt, image_input, image_strength, cfg_scale, steps, selected_ind
     selected_loras = [loras[idx] for idx in selected_indices]
 
     # Build the prompt with trigger words
-    prompt_mash = prompt
+    prepends = []
+    appends = []
     for lora in selected_loras:
         trigger_word = lora.get('trigger_word', '')
         if trigger_word:
             if lora.get("trigger_position") == "prepend":
-                prompt_mash = f"{trigger_word} {prompt_mash}"
+                prepends.append(trigger_word)
             else:
-                prompt_mash = f"{prompt_mash} {trigger_word}"
-
+                appends.append(trigger_word)
+    prompt_mash = " ".join(prepends + [prompt] + appends)
+    
     # Unload previous LoRA weights
     with calculateDuration("Unloading LoRA"):
         pipe.unload_lora_weights()
         pipe_i2i.unload_lora_weights()
 
     # Load LoRA weights with respective scales
+    lora_names = []
     with calculateDuration("Loading LoRA weights"):
         for idx, lora in enumerate(selected_loras):
+            lora_name = f"lora_{idx}"
+            lora_names.append(lora_name)
             lora_path = lora['repo']
             scale = lora_scale_1 if idx == 0 else lora_scale_2
             if image_input is not None:
                 if "weights" in lora:
-                    pipe_i2i.load_lora_weights(lora_path, weight_name=lora["weights"], multiplier=scale)
+                    pipe_i2i.load_lora_weights(lora_path, weight_name=lora["weights"], low_cpu_mem_usage=True, adapter_name=lora_name)
                 else:
-                    pipe_i2i.load_lora_weights(lora_path, multiplier=scale)
+                    pipe_i2i.load_lora_weights(lora_path, low_cpu_mem_usage=True, adapter_name=lora_name)
             else:
                 if "weights" in lora:
-                    pipe.load_lora_weights(lora_path, weight_name=lora["weights"], multiplier=scale)
+                    pipe.load_lora_weights(lora_path, weight_name=lora["weights"], low_cpu_mem_usage=True, adapter_name=lora_name)
                 else:
-                    pipe.load_lora_weights(lora_path, multiplier=scale)
+                    pipe.load_lora_weights(lora_path, low_cpu_mem_usage=True, adapter_name=lora_name)
 
+    pipeline.set_adapters(lora_names, adapter_weights=[lora_scale_1, lora_scale_2])
+    
     # Set random seed for reproducibility
     with calculateDuration("Randomizing seed"):
         if randomize_seed: