Fix layer_norm_eps in BERTEncoder (#1215)

src/gluonnlp/model/bert.py

@@ -318,11 +318,12 @@ def __init__(self, *, num_layers=2, units=512, hidden_size=2048,
         self._output_attention = output_attention
         self._output_all_encodings = output_all_encodings
         self._dropout = dropout
+        self._layer_norm_eps = layer_norm_eps
 
         with self.name_scope():
             if dropout:
                 self.dropout_layer = nn.Dropout(rate=dropout)
-            self.layer_norm = nn.LayerNorm(in_channels=units, epsilon=1e-12)
+            self.layer_norm = nn.LayerNorm(in_channels=units, epsilon=self._layer_norm_eps)
             self.position_weight = self.params.get('position_weight', shape=(max_length, units),
                                                    init=weight_initializer)
             self.transformer_cells = nn.HybridSequential()
@@ -550,7 +551,7 @@ def _get_decoder(self, units, vocab_size, embed, prefix):
             decoder = nn.HybridSequential(prefix=prefix)
             decoder.add(nn.Dense(units, flatten=False))
             decoder.add(GELU())
-            decoder.add(nn.LayerNorm(in_channels=units, epsilon=1e-12))
+            decoder.add(nn.LayerNorm(in_channels=units, epsilon=self.encoder._layer_norm_eps))
             decoder.add(nn.Dense(vocab_size, flatten=False, params=embed.collect_params()))
         assert decoder[3].weight == list(embed.collect_params().values())[0], \
             'The weights of word embedding are not tied with those of decoder'