Minor improvement while running custom models

modules/text_generation.py

@@ -97,6 +97,7 @@ def generate_reply(question, max_new_tokens, do_sample, temperature, top_p, typi
             print(f"Output generated in {(t1-t0):.2f} seconds.")
             yield formatted_outputs(reply, shared.model_name)
         else:
+            yield formatted_outputs(question, shared.model_name)
             for i in tqdm(range(max_new_tokens//8+1)):
                 clear_torch_cache()
                 reply = shared.model.generate(question, token_count=8, temperature=temperature, top_p=top_p)