diff --git a/tools/llm_bench/task/text_generation.py b/tools/llm_bench/task/text_generation.py
index 485de94996..4822b228ca 100644
--- a/tools/llm_bench/task/text_generation.py
+++ b/tools/llm_bench/task/text_generation.py
@@ -301,7 +301,7 @@ def token_printer():
         - np.array(perf_metrics.raw_metrics.m_new_token_times[:-1])
     ).tolist()
 
-    tm_list = np.array([first_token_time] + second_tokens_durations) / 1000
+    tm_list = (np.array([first_token_time] + second_tokens_durations) / 1000).tolist()
     inference_durations = (np.array(perf_metrics.raw_metrics.token_infer_durations) / 1000 / 1000).tolist()
     log.debug('latency of all tokens:')
     [log.debug('[{}]{:.4f}'.format(idx, tm)) for idx, tm in enumerate(tm_list)]
@@ -323,8 +323,8 @@ def token_printer():
     metrics_print.print_metrics(
         num,
         iter_data,
-        tm_list.tolist(),
-        inference_durations.tolist(),
+        tm_list,
+        inference_durations,
         warm_up=(num == 0),
         max_rss_mem=max_rss_mem_consumption,
         max_shared_mem=max_shared_mem_consumption,