diff --git a/run_once.sh b/run_once.sh
index 493c0f60..56a78b10 100644
--- a/run_once.sh
+++ b/run_once.sh
@@ -37,27 +37,27 @@ python prepare_yaml_file.py ${run_type} ${model} ${max_steps} ${run_name} ${outp
 
 export USE_MODELSCOPE_HUB=1
 
-# echo "Start recording npu status "
-# bash npu_status.sh ${output_dir} 60 0 &
-# npu_status_pid=$!
-# # echo "${npu_status_pid}"
+echo "Start recording npu status "
+bash npu_status.sh ${output_dir} 60 0 &
+npu_status_pid=$!
+# echo "${npu_status_pid}"
 
-# if [ "${gpu_cnt}" = "1" ]; then
-#    ASCEND_RT_VISIBLE_DEVICES=0 llamafactory-cli train ${output_dir}/${run_name}.yaml | tee "${output_dir}/log.txt" &
-#    train_pid=$!
-#    echo "Start single npu train"
-# else
-#    FORCE_TORCHRUN=1 llamafactory-cli train ${output_dir}/${run_name}.yaml | tee "${output_dir}/log.txt" &
-#    train_pid=$!
-#    echo "Start multi npu train"
-# fi
+if [ "${gpu_cnt}" = "1" ]; then
+   ASCEND_RT_VISIBLE_DEVICES=0 llamafactory-cli train ${output_dir}/${run_name}.yaml | tee "${output_dir}/log.txt" &
+   train_pid=$!
+   echo "Start single npu train"
+else
+   FORCE_TORCHRUN=1 llamafactory-cli train ${output_dir}/${run_name}.yaml | tee "${output_dir}/log.txt" &
+   train_pid=$!
+   echo "Start multi npu train"
+fi
 
-# wait $train_pid
-# echo "Train ended"
+wait $train_pid
+echo "Train ended"
 
-# sleep 60
-# kill $npu_status_pid
-# echo "Npu status ended"
+sleep 60
+kill $npu_status_pid
+echo "Npu status ended"
 
 python3 -c "import moxing as mox; mox.file.copy_parallel('${output_dir}', 'obs://xty/results/${run_name}')"
 rm -r ${output_dir}
\ No newline at end of file