fix #971
This commit is contained in:
@@ -427,7 +427,7 @@ XLM_INPUTS_DOCSTRING = r"""
|
||||
:func:`pytorch_transformers.PreTrainedTokenizer.convert_tokens_to_ids` for details.
|
||||
**position_ids**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
|
||||
Indices of positions of each input sequence tokens in the position embeddings.
|
||||
Selected in the range ``[0, config.max_position_embeddings - 1[``.
|
||||
Selected in the range ``[0, config.max_position_embeddings - 1]``.
|
||||
**token_type_ids**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
|
||||
A parallel sequence of tokens (can be used to indicate various portions of the inputs).
|
||||
The embeddings from these tokens will be summed with the respective token embeddings.
|
||||
|
||||
Reference in New Issue
Block a user