Update app.py

Make all common generate hyperparams parameterized ...

app.py

@@ -37,7 +37,7 @@ def build_messages(history, enable_thinking: bool):
     messages.insert(0, {"role": "system", "content": system_flag})
     return messages
 
-def chat_fn(history, enable_thinking, temperature, top_p, top_k, repetition_penalty):
+def chat_fn(history, enable_thinking, temperature, top_p, top_k, repetition_penalty, max_new_tokens):
     """Generate a streaming response."""
     messages = build_messages(history, enable_thinking)
     text = tokenizer.apply_chat_template(
@@ -50,7 +50,7 @@ def chat_fn(history, enable_thinking, temperature, top_p, top_k, repetition_pena
 
     streamer = model.generate(
         **inputs,
-        max_new_tokens=1024,
+        max_new_tokens=max_new_tokens,
         do_sample=True,
         temperature=temperature,
         top_p=top_p,
@@ -79,6 +79,7 @@ with gr.Blocks(title="SmolLM3-3B Chat") as demo:
         top_p = gr.Slider(0.0, 1.0, value=0.95, label="Top-p")
         top_k = gr.Slider(1,40,value=20,label="Top_k")
         repetition_penalty = gr.Slider(1.0,1.4,value=1.1,label="Repetition_Penalty")
+        max_new_tokens = gr.Slider(1000,32768,value=32768,label="Max_New_Tokens")
     chatbot = gr.Chatbot(type="messages")
     msg = gr.Textbox(placeholder="Type your message here…", lines=1)
     clear = gr.Button("Clear")