Update transformer_mt_roberta/modeling_transformer_final.py

transformer_mt_roberta/modeling_transformer_final.py

@@ -8,7 +8,7 @@ import torch.nn.functional as F
 
 from transformer_mt.modeling_attention import MultiHeadAttention
 from transformer_mt.utils import pad
-from transformers import AutoTokenizer, AutoModelForMaskedML
+from transformers import AutoTokenizer, AutoModelForMaskedLM
 
 Hypothesis = namedtuple("Hypothesis", ["value", "score"])
 
@@ -91,7 +91,7 @@ class TransfomerEncoderDecoderModel(nn.Module):
 
         self.dropout = nn.Dropout(self.dropout_rate)
 
-        self.encoder = AutoModelForMaskedML.from_pretrained("flax-community/roberta_base_danish", output_hidden_states=True)
+        self.encoder = AutoModelForMaskedLM.from_pretrained("flax-community/roberta_base_danish", output_hidden_states=True)
 
         self.decoder_layers = nn.ModuleList([TransformerDecoderLayer(hidden = self.hidden,
                                                                     num_heads = self.num_heads,