Update app.py

app.py

@@ -55,9 +55,7 @@ tokenizer, model, image_processor, context_len = load_pretrained_model(
 model.config.training = False
 
 # FIXED PROMPT
-FIXED_PROMPT = "What type of waste is this item and how to dispose of it?"
-INTERNAL_PROMPT = "<image>\nWhat type of waste is this item and how to dispose of it?"
-
+FIXED_PROMPT = "<image>\nWhat type of waste is this item and how to dispose of it?"
 
 def clear_history():
     state = default_conversation.copy()
@@ -68,15 +66,13 @@ def add_text(state, imagebox, textbox, image_process_mode):
         state = conv_templates[conv_mode].copy()
 
     if imagebox is not None:
+        textbox = FIXED_PROMPT
         image = Image.open(imagebox).convert('RGB')
-        visible_text = FIXED_PROMPT
-        model_prompt = (INTERNAL_PROMPT, image, image_process_mode)
-        state.append_message(state.roles[0], visible_text)
-        state.append_message(state.roles[1], None)
-        state.messages[-2].append(model_prompt)  
+        textbox = (textbox, image, image_process_mode)
+    state.append_message(state.roles[0], textbox)
+    state.append_message(state.roles[1], None)
     yield (state, state.to_gradio_chatbot(), "", None) + (disable_btn, disable_btn, disable_btn, enable_btn, enable_btn)
 
-
 def delete_text(state, image_process_mode):
     state.messages[-1][-1] = None
     prev_human_msg = state.messages[-2]
@@ -94,11 +90,7 @@ def regenerate(state, image_process_mode):
 
 @spaces.GPU
 def generate(state, imagebox, textbox, image_process_mode, temperature, top_p, max_output_tokens):
-    hidden_data = state.messages[-2]
-    if len(hidden_data) == 3 and isinstance(hidden_data[2], tuple):
-        prompt, image, image_process_mode = hidden_data[2]
-    else:
-        prompt = FIXED_PROMPTT
+    prompt = FIXED_PROMPT
     images = state.get_images(return_pil=True)
 
     ori_prompt = prompt
@@ -175,10 +167,8 @@ If you use this service or build upon this work, please cite the original public
 Li, Jiachen and Wang, Xinyao and Zhu, Sijie and Kuo, Chia-wen and Xu, Lu and Chen, Fan and Jain, Jitesh and Shi, Humphrey and Wen, Longyin.  
 CuMo: Scaling Multimodal LLM with Co-Upcycled Mixture-of-Experts. arXiv preprint, 2024.  
 [[arXiv](https://arxiv.org/abs/2405.05949)]
-
 By using this service, users are required to agree to the following terms:  
 The service is a research preview intended for non-commercial use only. It only provides limited safety measures and may generate offensive content. It must not be used for any illegal, harmful, violent, racist, or sexual purposes.
-
 For an optimal experience, please use desktop computers for this demo, as mobile devices may compromise its quality.
 """)
 
@@ -199,7 +189,7 @@ block_css = """
 
 textbox = gr.Textbox(
     show_label=False, 
-    placeholder="Prompt is fixed: What material is this item and how to dispose of it?", 
+    placeholder="Prompt is fixed: What type of waste is this item and how to dispose of it?", 
     container=False,
     interactive=False
 )
@@ -220,7 +210,7 @@ with gr.Blocks(title="CuMo", theme=gr.themes.Default(), css=block_css) as demo:
     
             #cur_dir = os.path.dirname(os.path.abspath(__file__))
             cur_dir = './cumo/serve'
-            default_prompt = "<image>\nWhat material is this item and how to dispose of it?"
+            default_prompt = "<image>\nWhat type of waste is this item and how to dispose of it?"
             gr.Examples(examples=[
                 [f"{cur_dir}/examples/0165 CB.jpg", default_prompt],
                 [f"{cur_dir}/examples/0225 PA.jpg", default_prompt],