Update

app.py

@@ -40,7 +40,7 @@ def get_model_response(
         model, adapter_name, processor, index_to_token, role: str, 
         image_paths: List[str], user_message: str = "", target_image: str = ""
 ) -> str:    
-    #model.model.set_adapter(adapter_name)
+    model.model.set_adapter(adapter_name)
     if role == "speaker":
         img_dir = "tangram_pngs"
         print("Starting processing")
@@ -69,7 +69,7 @@ def get_model_response(
 
     return response
 
-@spaces.GPU(duration=15)
+@spaces.GPU(duration=20)
 def get_speaker_response(model, images, input_tokens, attn_mask, image_attn_mask, label, image_paths, processor, img_dir, index_to_token):
     model = model.cuda()
     with torch.no_grad():
@@ -81,7 +81,7 @@ def get_speaker_response(model, images, input_tokens, attn_mask, image_attn_mask
         )
     return captions
 
-@spaces.GPU(duration=15)
+@spaces.GPU(duration=20)
 def get_listener_response(model, images, l_input_tokens, l_attn_mask, l_image_attn_mask, index_to_token,
                           s_input_tokens, s_attn_mask, s_image_attn_mask, s_target_mask, s_target_label, image_paths):
     model = model.cuda()