Add detailed log for llm service

PaddlePaddle · Dec 14, 2023 · 345f788 · 345f788
1 parent 73b5af2
commit 345f788
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/llm/fastdeploy_llm/serving/triton_model.py b/llm/fastdeploy_llm/serving/triton_model.py
@@ -45,8 +45,8 @@ def stream_call_back(call_back_task, token_tuple, index, is_last_token,
         "OUT", np.array(
             [json.dumps(out)], dtype=np.object_))
     if is_last_token:
-        all_token_ids = [t[0] for t in call_back_task.result.completion_tokens]
-        all_strs = "".join[t[1] for t in call_back_task.result.completion_tokens]   
+        all_token_ids = [t[0] for t in call_back_task.result.completion_tokens] + [token_tuple[0]]
+        all_strs = "".join[t[1] for t in call_back_task.result.completion_tokens] + [token_tuple[1]] 
         logger.info("Model output for req_id: {}  results_all: {} tokens_all: {}".format(call_back_task.task_id, all_strs, all_token_ids))
         sender[call_back_task.task_id].send(
             pb_utils.InferenceResponse([out_tensor]),