revert bad changes

src/transformers/modeling_utils.py

@@ -1643,11 +1643,11 @@ def _set_default_torch_dtype(cls, dtype: torch.dtype) -> torch.dtype:
  return dtype_orig
 
  @property
- def base_model(self):
+ def base_model(self) -> nn.Module:
  """
  `torch.nn.Module`: The main body of the model.
  """
- return getattr(self, "base_model_prefix", self)
+ return getattr(self, self.base_model_prefix, self)
 
  @classmethod
  def can_generate(cls) -> bool:

src/transformers/models/deberta/modeling_deberta.py

@@ -930,7 +930,12 @@ def forward(self, sequence_output, word_embeddings):
 
 @add_start_docstrings("""DeBERTa Model with a `language modeling` head on top.""", DEBERTA_START_DOCSTRING)
 class DebertaForMaskedLM(DebertaPreTrainedModel):
- _tied_weights_keys = ["cls.predictions.decoder.weight", "cls.predictions.decoder.bias"]
+ _tied_weights_keys = [
+ "cls.predictions.decoder.weight",
+ "cls.predictions.decoder.bias",
+ "deberta.embeddings.word_embeddings.weight",
+ "lm_predictions.lm_head.weight",
+ ]
 
  def __init__(self, config):
  super().__init__(config)

src/transformers/trainer.py

@@ -3512,8 +3512,8 @@ def training_step(self, model: nn.Module, inputs: Dict[str, Union[torch.Tensor,
  `torch.Tensor`: The tensor with training loss on this batch.
  """
  model.train()
- # if hasattr(self.optimizer, "train") and callable(self.optimizer.train):
- #  self.optimizer.train()
+ if hasattr(self.optimizer, "train") and callable(self.optimizer.train):
+ self.optimizer.train()
 
  inputs = self._prepare_inputs(inputs)
  if is_sagemaker_mp_enabled():