Update modeling_norbert.py

modeling_norbert.py

@@ -57,7 +57,6 @@ class MaskClassifier(nn.Module):
             nn.Dropout(config.hidden_dropout_prob),
             nn.Linear(subword_embedding.size(1), subword_embedding.size(0))
         )
-        self.initialize(config.hidden_size, subword_embedding)
 
     def forward(self, x, masked_lm_labels=None):
         if masked_lm_labels is not None:
@@ -97,7 +96,6 @@ class FeedForward(nn.Module):
             nn.Linear(config.intermediate_size, config.hidden_size, bias=False),
             nn.Dropout(config.hidden_dropout_prob)
         )
-        self.initialize(config.hidden_size)
 
     def forward(self, x):
         return self.mlp(x)
@@ -148,7 +146,6 @@ class Attention(nn.Module):
 
         self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
         self.scale = 1.0 / math.sqrt(3 * self.head_size)
-        self.initialize()
 
     def make_log_bucket_position(self, relative_pos, bucket_size, max_position):
         sign = torch.sign(relative_pos)