Fix the masked LM model
This commit is contained in:
@@ -925,6 +925,9 @@ class TapasForMaskedLM(TapasPreTrainedModel):
|
||||
def get_output_embeddings(self):
|
||||
return self.lm_head
|
||||
|
||||
def set_output_embeddings(self, word_embeddings):
|
||||
self.lm_head = word_embeddings
|
||||
|
||||
@add_start_docstrings_to_model_forward(TAPAS_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
|
||||
@replace_return_docstrings(output_type=MaskedLMOutput, config_class=_CONFIG_FOR_DOC)
|
||||
def forward(
|
||||
|
||||
Reference in New Issue
Block a user