update app

app.py

@@ -80,6 +80,7 @@ def generate_stream(prompt, max_new_tokens=128):
             model.generate(
                 **inputs,
                 max_new_tokens=max_new_tokens,
+                min_new_tokens=16,
                 do_sample=True,
                 temperature=0.7,
                 top_p=0.9,