diff --git a/egs/aishell/ASR/transformer_ctc/.decode.py.swp b/egs/aishell/ASR/transformer_ctc/.decode.py.swp
index e80ac848b..d53020f65 100644
Binary files a/egs/aishell/ASR/transformer_ctc/.decode.py.swp and b/egs/aishell/ASR/transformer_ctc/.decode.py.swp differ
diff --git a/egs/aishell/ASR/transformer_ctc/decode.py b/egs/aishell/ASR/transformer_ctc/decode.py
index 4fc8ecb26..f69128454 100755
--- a/egs/aishell/ASR/transformer_ctc/decode.py
+++ b/egs/aishell/ASR/transformer_ctc/decode.py
@@ -453,6 +453,12 @@ def decode_dataset(
         num_batches = "?"
 
     results = defaultdict(list)
+
+    subs_all = 0
+    dels_all = 0
+    ins_all = 0
+    char_num = 0
+
     for batch_idx, batch in enumerate(dl):
         #logging.info(f"decoding {batch_idx} th batch")
         texts = batch["supervisions"]["text"]
@@ -483,9 +489,18 @@ def decode_dataset(
         )
 
         for i, hyp in enumerate(hyps):
-            print('hyp = ', hyp)
-            print('ref = ', texts[i].replace(' ', ''))
-            print('')
+            #print('hyp = ', hyp)
+            #print('ref = ', texts[i].replace(' ', ''))
+            #print('')
+            ref = texts[i].replace(' ', '')
+            [cer, subs, dels, ins] = metrics.get_cer(hyp, ref)
+            subs_all += subs
+            dels_all += dels
+            ins_all += ins
+            char_num == len(ref)
+    
+    cer = (subs_all+dels_all+ins_all) / char_num
+    print(cer * 100)
         
         '''
         for lm_scale, hyps in hyps_dict.items():