Update app.py

app.py

@@ -69,12 +69,12 @@ PLACEHOLDER = """
 def respond(
     message,
     history: list[tuple[str, str]],
+    model,
     max_tokens,
     temperature,
     top_p,
     top_k,
     repeat_penalty,
-    model,
 ):  
     llm = Llama(
         model_path=f"models/{model}",
@@ -125,6 +125,7 @@ def respond(
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
+        gr.Dropdown(['dolphin-2.9.1-yi-1.5-34b-Q6_K.gguf', 'dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf'], value="dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf", label="Model"),
         gr.Slider(minimum=1, maximum=8192, value=8192, step=1, label="Max tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(
@@ -148,7 +149,6 @@ demo = gr.ChatInterface(
             step=0.1,
             label="Repetition penalty",
         ),
-        gr.Dropdown(['dolphin-2.9.1-yi-1.5-34b-Q6_K.gguf', 'dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf'], value="dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf", label="Model"),
     ],
     theme=gr.themes.Soft(primary_hue="indigo", secondary_hue="blue", neutral_hue="gray",font=[gr.themes.GoogleFont("Exo"), "ui-sans-serif", "system-ui", "sans-serif"]).set(
         body_background_fill_dark="#0f172a",