Update app.py

app.py

@@ -137,11 +137,10 @@ templates = [
     "B22", "LLAMA_3", "PHI_3"
 ]
 
-with gr.Blocks() as demo:
-    gr.Markdown(description)
-    with gr.Row():
-        model_dropdown = gr.Dropdown(
-            choices=[
+demo = gr.ChatInterface(
+    fn=respond,
+    inputs=[
+        gr.Dropdown([
                 'Oumuamua-7b-RP_Q4_K_M.gguf',
                 'Oumuamua-7b-instruct-v2-Q4_K_M.gguf',
                 'umiyuki-Umievo-itr012-Gleipnir-7B-Q4_K_M.gguf',
@@ -151,47 +150,49 @@ with gr.Blocks() as demo:
             ],
             value="Oumuamua-7b-RP_Q4_K_M.gguf",
             label="Model"
-        )
-        template_dropdown = gr.Dropdown(
+        ),
+        gr.Dropdown(
             choices=templates,
             value="LLAMA_3",
             label="Template"
-        )
+        ),
+        gr.Textbox(value="You are a helpful assistant.", label="System message"),
+        gr.Slider(minimum=1, maximum=4096, value=2048, step=1, label="Max tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p",
+        ),
+        gr.Slider(
+            minimum=0,
+            maximum=100,
+            value=40,
+            step=1,
+            label="Top-k",
+        ),
+        gr.Slider(
+            minimum=0.0,
+            maximum=2.0,
+            value=1.1,
+            step=0.1,
+            label="Repetition penalty",
+        ),
+    ],
+    retry_btn="Retry",
+    undo_btn="Undo",
+    clear_btn="Clear",
+    submit_btn="Send",
+    title="Chat with various models using llama.cpp", 
+    description=description,
+    chatbot=gr.Chatbot(
+        scale=1, 
+        likeable=False,
+        show_copy_button=True
+    )
+)
 
-    chatbot = gr.Chatbot()
-    with gr.Row():
-        system_message = gr.Textbox(value="You are a helpful assistant.", label="System message")
-    with gr.Row():
-        max_tokens = gr.Slider(minimum=1, maximum=4096, value=2048, step=1, label="Max tokens")
-        temperature = gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature")
-    with gr.Row():
-        top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p")
-        top_k = gr.Slider(minimum=0, maximum=100, value=40, step=1, label="Top-k")
-    with gr.Row():
-        repeat_penalty = gr.Slider(minimum=0.0, maximum=2.0, value=1.1, step=0.1, label="Repetition penalty")
-
-    submit_btn = gr.Button("Send")
-    clear_btn = gr.Button("Clear")
-    undo_btn = gr.Button("Undo")
-    retry_btn = gr.Button("Retry")
-
-    def handle_response(message, history):
-        return list(respond(
-            message,
-            history,
-            model_dropdown.value,
-            system_message.value,
-            max_tokens.value,
-            temperature.value,
-            top_p.value,
-            top_k.value,
-            repeat_penalty.value,
-            template_dropdown.value
-        ))
-
-    submit_btn.click(fn=handle_response, inputs=[chatbot.input, chatbot.history], outputs=chatbot)
-    clear_btn.click(lambda: None, None, chatbot)
-    undo_btn.click(lambda x: x[:-1] if x else None, chatbot, chatbot)
-    retry_btn.click(lambda x: x[:-1] if x else None, chatbot, chatbot)
-
-demo.launch()
+if __name__ == "__main__":
+    demo.launch()