Do sample

app.py

@@ -74,7 +74,7 @@ with gr.Blocks() as demo:
         for token_id in current_model.generate(
             **inputs,
             max_new_tokens=256,
-            do_sample=False,
+            do_sample=True,
             return_dict_in_generate=True,
             output_scores=False
         ).sequences[0][inputs['input_ids'].shape[-1]:]:  # skip input tokens