Update app.py

app.py

@@ -86,7 +86,7 @@ def load_b_lora_to_unet(pipe, content_lora_model_id: str = '', style_lora_model_
         except Exception as e:
             raise type(e)(f'failed to load_b_lora_to_unet, due to: {e}')
 
-@spaces.GPU()
+
 def load_b_loras(content_b_lora, style_b_lora):
     
     if content_b_lora != "" and content_b_lora is not None:
@@ -108,7 +108,16 @@ def load_b_loras(content_b_lora, style_b_lora):
 
     prepared_prompt = f"{content_model_instance_prompt} {style_model_instance_prompt}"
 
-    pipeline.unload_lora_weights()
+    
+    return prepared_prompt
+
+@spaces.GPU()
+def main(content_b_lora, style_b_lora, prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):
+
+    if randomize_seed:
+        seed = random.randint(0, MAX_SEED)
+        
+    generator = torch.Generator().manual_seed(seed)
     
     if content_b_lora is None:
         content_B_LoRA_path = ''
@@ -124,16 +133,6 @@ def load_b_loras(content_b_lora, style_b_lora):
 
     load_b_lora_to_unet(pipeline, content_B_LoRA_path, style_B_LoRA_path, content_alpha, style_alpha)
     
-    return prepared_prompt
-
-@spaces.GPU()
-def main(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):
-
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-        
-    generator = torch.Generator().manual_seed(seed)
-    
     prompt = prompt
     image = pipeline(
         prompt,
@@ -142,7 +141,8 @@ def main(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_
         width = width, 
         height = height,
     ).images[0]
-    
+
+    pipeline.unload_lora_weights()
     return image
 
 css="""
@@ -257,7 +257,7 @@ with gr.Blocks(css=css) as demo:
     
     run_button.click(
         fn = main,
-        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
+        inputs = [content_b_lora, style_b_lora, prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
         outputs = [result]
     )