Update app.py

app.py

@@ -11,7 +11,7 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 phi4_model = AutoModelForCausalLM.from_pretrained(phi4_model_path, device_map="auto", torch_dtype="auto")
 phi4_tokenizer = AutoTokenizer.from_pretrained(phi4_model_path)
 
-@spaces.GPU(duration=60)
+@spaces.GPU(duration=120)
 def generate_response(user_message, max_tokens, temperature, top_k, top_p, repetition_penalty, history_state):
     if not user_message.strip():
         return history_state, history_state
@@ -133,7 +133,7 @@ with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
                 repetition_penalty_slider = gr.Slider(1.0, 2.0, value=1.0, label="Repetition Penalty")
 
         with gr.Column(scale=4):
-            chatbot = gr.Chatbot(label="Chat", render_markdown=True, type="messages", show_copy_button=True)
+            chatbot = gr.Chatbot(label="Chat", render_markdown=True, show_copy_button=True)
             with gr.Row():
                 user_input = gr.Textbox(label="Describe symptoms or ask a medical question", placeholder="Type your message here...", scale=3)
                 submit_button = gr.Button("Send", variant="primary", scale=1)
@@ -145,8 +145,17 @@ with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
                 example3 = gr.Button("Abdominal pain")
                 example4 = gr.Button("BMI calculation")
 
-    submit_button.stream(
-        fn=generate_response,
+    # Replace the streaming approach with a standard click event
+    def process_input(message, max_tokens, temperature, top_k, top_p, repetition_penalty, history):
+        # Call the generator function and get the final result
+        generator = generate_response(message, max_tokens, temperature, top_k, top_p, repetition_penalty, history)
+        result = None
+        for res in generator:
+            result = res
+        return result
+
+    submit_button.click(
+        fn=process_input,
         inputs=[user_input, max_tokens_slider, temperature_slider, top_k_slider, top_p_slider, repetition_penalty_slider, history_state],
         outputs=[chatbot, history_state]
     ).then(
@@ -162,4 +171,4 @@ with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
     example3.click(lambda: gr.update(value=example_messages["Abdominal pain"]), None, user_input)
     example4.click(lambda: gr.update(value=example_messages["BMI calculation"]), None, user_input)
 
-demo.launch(ssr_mode=False)
+demo.launch(ssr_mode=False)