diff --git a/PaddleNLP/ELMO/args.py b/PaddleNLP/ELMO/args.py
index 53bc0191280718d5ab44b8d482495b335d5d1e96..8653cc0315b166f84f3d5238e9afa9b470a57120 100755
--- a/PaddleNLP/ELMO/args.py
+++ b/PaddleNLP/ELMO/args.py
@@ -94,7 +94,7 @@ def parse_args():
     parser.add_argument('--cell_clip', type=float, default=3.0)
     parser.add_argument('--max_epoch', type=float, default=10)
     parser.add_argument('--local', type=bool, default=False)
-    parser.add_argument('--shuffle', type=bool, default=False)
+    parser.add_argument('--shuffle', type=bool, default=True)
     parser.add_argument('--use_custom_samples', type=bool, default=False)
     parser.add_argument('--para_save_dir', type=str, default='model_new')
     parser.add_argument('--train_path', type=str, default='')
diff --git a/PaddleNLP/ELMO/run.sh b/PaddleNLP/ELMO/run.sh
index 1edf909738ce2664887e3dc0a9cbd9461a748dd2..888f0203b74c41f4d5a41cfe3ea1c1967980bf40 100755
--- a/PaddleNLP/ELMO/run.sh
+++ b/PaddleNLP/ELMO/run.sh
@@ -5,4 +5,5 @@ python  train.py \
 --vocab_path baike/vocabulary_min5k.txt \
 --learning_rate 0.2 \
 --use_gpu True \
+--shuffle True \
 --local True $@
diff --git a/PaddleNLP/ELMO/train.py b/PaddleNLP/ELMO/train.py
index 4f4a346eceec558144bb4b411128cd8eb6e5b787..584fb1190d9a421a07b6155d6f708cfc3052198b 100755
--- a/PaddleNLP/ELMO/train.py
+++ b/PaddleNLP/ELMO/train.py
@@ -578,7 +578,6 @@ def train_loop(args,
 
         end_time = time.time()
         total_time += end_time - start_time
-        logger.info("train ppl {}".format(ppl))
 
         if epoch_id == args.max_epoch - 1 and args.enable_ce:
             logger.info("lstm_language_model_duration\t%s" %