diff --git a/egs/aishell/ASR/conformer_ctc/.transformer.py.swp b/egs/aishell/ASR/conformer_ctc/.transformer.py.swp
index 08932d237..0ea38276f 100644
Binary files a/egs/aishell/ASR/conformer_ctc/.transformer.py.swp and b/egs/aishell/ASR/conformer_ctc/.transformer.py.swp differ
diff --git a/egs/aishell/ASR/conformer_ctc/transformer.py b/egs/aishell/ASR/conformer_ctc/transformer.py
index b82cc486f..d70abb6d2 100644
--- a/egs/aishell/ASR/conformer_ctc/transformer.py
+++ b/egs/aishell/ASR/conformer_ctc/transformer.py
@@ -23,8 +23,6 @@ import torch.nn as nn
 from label_smoothing import LabelSmoothingLoss
 from subsampling import Conv2dSubsampling, VggSubsampling
 from torch.nn.utils.rnn import pad_sequence
-from torch.nn.modules import Module
-from torch import Tensor
 # Note: TorchScript requires Dict/List/etc. to be fully typed.
 Supervisions = Dict[str, torch.Tensor]
 
@@ -382,7 +380,7 @@ class Transformer(nn.Module):
         return nll
 
 
-class TransformerEncoder(Module):
+class TransformerEncoder(nn.TransformerEncoder):
     r"""TransformerEncoder is a stack of N encoder layers. Users can build the
     BERT(https://arxiv.org/abs/1810.04805) model with corresponding parameters.