The XLM-RoBERTa model was proposed in `Unsupervised Cross-lingual Representation Learning at Scale`_
by Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov. It is based on Facebook's RoBERTa model released in 2019.
It is a large multi-lingual language model, trained on 2.5TB of filtered CommonCrawl data.
This implementation is the same as RoBERTa.
This model is a PyTorch `torch.nn.Module`_ sub-class. Use it as a regular PyTorch Module and
refer to the PyTorch documentation for all matter related to general usage and behavior.
.. _`Unsupervised Cross-lingual Representation Learning at Scale`:
XLM_ROBERTA_START_DOCSTRING=r""" The XLM-RoBERTa model was proposed in
XLM_ROBERTA_START_DOCSTRING=r"""
`Unsupervised Cross-lingual Representation Learning at Scale`_
by Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov. It is based on Facebook's RoBERTa model released in 2019.
It is a large multi-lingual language model, trained on 2.5TB of filtered CommonCrawl data.
This implementation is the same as RoBERTa.
This model is a PyTorch `torch.nn.Module`_ sub-class. Use it as a regular PyTorch Module and
This model is a PyTorch `torch.nn.Module`_ sub-class. Use it as a regular PyTorch Module and
refer to the PyTorch documentation for all matter related to general usage and behavior.
refer to the PyTorch documentation for all matter related to general usage and behavior.
.. _`Unsupervised Cross-lingual Representation Learning at Scale`:
https://arxiv.org/abs/1911.02116
.. _`torch.nn.Module`:
https://pytorch.org/docs/stable/nn.html#module
Parameters:
Parameters:
config (:class:`~transformers.XLMRobertaConfig`): Model configuration class with all the parameters of the
config (:class:`~transformers.XLMRobertaConfig`): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the configuration.
model. Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""
"""
XLM_ROBERTA_INPUTS_DOCSTRING=r"""
Inputs:
**input_ids**: ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Indices of input sequence tokens in the vocabulary.
To match pre-training, XLM-RoBERTa input sequence should be formatted with <s> and </s> tokens as follows:
(a) For sequence pairs:
``tokens: <s> Is this Jacksonville ? </s> </s> No it is not . </s>``
(b) For single sequences:
``tokens: <s> the dog is hairy . </s>``
Fully encoded sequences or sequence pairs can be obtained using the XLMRobertaTokenizer.encode function with
the ``add_special_tokens`` parameter set to ``True``.
XLM-RoBERTa is a model with absolute position embeddings so it's usually advised to pad the inputs on
the right rather than the left.
See :func:`transformers.PreTrainedTokenizer.encode` and
:func:`transformers.PreTrainedTokenizer.convert_tokens_to_ids` for details.
**attention_mask**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length)``:
Mask to avoid performing attention on padding token indices.
Mask values selected in ``[0, 1]``:
``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
**token_type_ids**: (`optional` need to be trained) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
Optional segment token indices to indicate first and second portions of the inputs.
This embedding matrice is not trained (not pretrained during XLM-RoBERTa pretraining), you will have to train it
during finetuning.
Indices are selected in ``[0, 1]``: ``0`` corresponds to a `sentence A` token, ``1``
corresponds to a `sentence B` token
(see `BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding`_ for more 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[``.
**head_mask**: (`optional`) ``torch.FloatTensor`` of shape ``(num_heads,)`` or ``(num_layers, num_heads)``:
Mask to nullify selected heads of the self-attention modules.
Mask values selected in ``[0, 1]``:
``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
**inputs_embeds**: (`optional`) ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, embedding_dim)``:
Optionally, instead of passing ``input_ids`` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix.
"""
@add_start_docstrings(
@add_start_docstrings(
"The bare XLM-RoBERTa Model transformer outputting raw hidden-states without any specific head on top.",
"The bare XLM-RoBERTa Model transformer outputting raw hidden-states without any specific head on top.",
XLM_ROBERTA_START_DOCSTRING,
XLM_ROBERTA_START_DOCSTRING,
XLM_ROBERTA_INPUTS_DOCSTRING,
)
)
classXLMRobertaModel(RobertaModel):
classXLMRobertaModel(RobertaModel):
r"""
"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
This class overrides :class:`~transformers.RobertaModel`. Please check the
**last_hidden_state**: ``torch.FloatTensor`` of shape ``(batch_size, sequence_length, hidden_size)``
superclass for the appropriate documentation alongside usage examples.
Sequence of hidden-states at the output of the last layer of the model.
**pooler_output**: ``torch.FloatTensor`` of shape ``(batch_size, hidden_size)``
Last layer hidden-state of the first token of the sequence (classification token)
further processed by a Linear layer and a Tanh activation function. The Linear
layer weights are trained from the next sentence prediction (classification)
eo match pre-training, XLM-RoBERTa input sequence should be formatted with <s> and </s> tokens as follows:
(a) For sequence pairs:
``tokens: <s> is this jack ##son ##ville ? </s> </s> no it is not . </s>``
``token_type_ids: 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1``
(b) For single sequences:
``tokens: <s> the dog is hairy . </s>``
``token_type_ids: 0 0 0 0 0 0 0``
objective during Bert pretraining. This output is usually *not* a good summary
of the semantic content of the input, you're often better with averaging or pooling
the sequence of hidden-states for the whole input sequence.
**hidden_states**: (`optional`, returned when ``config.output_hidden_states=True``)
list of ``torch.FloatTensor`` (one for the output of each layer + the output of the embeddings)
of shape ``(batch_size, sequence_length, hidden_size)``:
Hidden-states of the model at the output of each layer plus the initial embedding outputs.
**attentions**: (`optional`, returned when ``config.output_attentions=True``)
list of ``torch.FloatTensor`` (one for each layer) of shape ``(batch_size, num_heads, sequence_length, sequence_length)``:
Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.
