@@ -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)