Fix missing eps arg for LayerNorm in ElectraGeneratorPredictions (#15332)

* fix missing eps

* Same fix for ConvBertGeneratorPredictions

* Same fix for AlbertMLMHead
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/albert/modeling_albert.py

@@ -856,7 +856,7 @@ class AlbertMLMHead(nn.Module):
     def __init__(self, config):
         super().__init__()
 
-        self.LayerNorm = nn.LayerNorm(config.embedding_size)
+        self.LayerNorm = nn.LayerNorm(config.embedding_size, eps=config.layer_norm_eps)
         self.bias = nn.Parameter(torch.zeros(config.vocab_size))
         self.dense = nn.Linear(config.hidden_size, config.embedding_size)
         self.decoder = nn.Linear(config.embedding_size, config.vocab_size)

src/transformers/models/convbert/modeling_convbert.py

@@ -865,7 +865,7 @@ class ConvBertGeneratorPredictions(nn.Module):
     def __init__(self, config):
         super().__init__()
 
-        self.LayerNorm = nn.LayerNorm(config.embedding_size)
+        self.LayerNorm = nn.LayerNorm(config.embedding_size, eps=config.layer_norm_eps)
         self.dense = nn.Linear(config.hidden_size, config.embedding_size)
 
     def forward(self, generator_hidden_states):

src/transformers/models/electra/modeling_electra.py

@@ -647,7 +647,7 @@ class ElectraGeneratorPredictions(nn.Module):
     def __init__(self, config):
         super().__init__()
 
-        self.LayerNorm = nn.LayerNorm(config.embedding_size)
+        self.LayerNorm = nn.LayerNorm(config.embedding_size, eps=config.layer_norm_eps)
         self.dense = nn.Linear(config.hidden_size, config.embedding_size)
 
     def forward(self, generator_hidden_states):