modeling_xlmr_decoupled.py

# coding=utf-8
# Copyright 2019 Facebook AI Research and the HuggingFace Inc. team.
# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""PyTorch XLM-RoBERTa model."""

import math
from typing import List, Optional, Tuple, Union

import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss

from transformers.activations import ACT2FN, gelu
from transformers.modeling_outputs import (
    BaseModelOutputWithPastAndCrossAttentions,
    BaseModelOutputWithPoolingAndCrossAttentions,
    CausalLMOutputWithCrossAttentions,
    MaskedLMOutput,
    MultipleChoiceModelOutput,
    QuestionAnsweringModelOutput,
    SequenceClassifierOutput,
    TokenClassifierOutput,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from transformers.utils import (
    add_code_sample_docstrings,
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    logging,
    replace_return_docstrings,
)

from transformers.models.xlm_roberta.modeling_xlm_roberta import (
    XLMRobertaEncoder,
    XLMRobertaPooler,
    XLMRobertaPreTrainedModel,
    XLMRobertaClassificationHead
)

logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "FacebookAI/xlm-roberta-base"
_CONFIG_FOR_DOC = "XLMRobertaConfig"


# Copied from transformers.models.roberta.modeling_roberta.RobertaEmbeddings with Roberta->XLMRoberta
class XLMRobertaDecoupledEmbeddings(nn.Module):
    """
    Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
    """

    # Copied from transformers.models.bert.modeling_bert.BertEmbeddings.__init__
    def __init__(self, config):
        super().__init__()
        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
        # self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)

        # this is used as the language embedding
        if config.use_lang_embedding is True:
            self.lang_type_embeddings = nn.Embedding(config.lang_size, config.hidden_size)

        # this is used as the script embedding
        if config.use_script_embedding is True:
            self.script_type_embeddings = nn.Embedding(config.script_size, config.hidden_size)

        # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
        # any TensorFlow checkpoint file
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
        self.register_buffer(
            "position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
        )
        self.register_buffer(
            "token_lang_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
        )
        self.register_buffer(
            "token_script_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
        )

        # End copy
        self.padding_idx = config.pad_token_id
        self.position_embeddings = nn.Embedding(
            config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
        )
        self.config = config

    def forward(
            self, token_lang_ids=None, token_script_ids=None,
            input_ids=None, position_ids=None, inputs_embeds=None, past_key_values_length=0
    ):
        if position_ids is None:
            if input_ids is not None:
                # Create the position ids from the input token ids. Any padded tokens remain padded.
                position_ids = create_position_ids_from_input_ids(input_ids, self.padding_idx, past_key_values_length)
            else:
                position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)

        if inputs_embeds is None:
            inputs_embeds = self.word_embeddings(input_ids)

        if self.config.decouple_at_input_embeddings:
            if token_lang_ids is None and self.config.use_lang_embedding is True:
                raise ValueError("token_lang_ids cannot be None if use_lang_embed is True")
            else:
                if self.config.use_lang_embedding is True:
                    token_lang_embeddings = self.lang_type_embeddings(token_lang_ids)
                    inputs_embeds = inputs_embeds + token_lang_embeddings
            if token_script_ids is None and self.config.use_script_embedding is True:
                raise ValueError("token_script_ids cannot be None if use_script_embedding is True")
            else:
                if self.config.use_script_embedding:
                    token_script_embeddings = self.script_type_embeddings(token_lang_ids)
                    inputs_embeds = inputs_embeds + token_script_embeddings

        embeddings = inputs_embeds

        if self.position_embedding_type == "absolute":
            position_embeddings = self.position_embeddings(position_ids)
            embeddings += position_embeddings
        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings

    def create_position_ids_from_inputs_embeds(self, inputs_embeds):
        """
        We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.

        Args:
            inputs_embeds: torch.Tensor

        Returns: torch.Tensor
        """
        input_shape = inputs_embeds.size()[:-1]
        sequence_length = input_shape[1]

        position_ids = torch.arange(
            self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
        )
        return position_ids.unsqueeze(0).expand(input_shape)


XLM_ROBERTA_START_DOCSTRING = r"""

    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
    etc.)

    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
    and behavior.

    Parameters:
        config ([`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. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
"""

XLM_ROBERTA_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (`torch.LongTensor` of shape `({0})`):
            Indices of input sequence tokens in the vocabulary.

            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.

            [What are input IDs?](../glossary#input-ids)
        attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:

            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.

            [What are attention masks?](../glossary#attention-mask)
        token_type_ids (`torch.LongTensor` of shape `({0})`, *optional*):
            Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0,
            1]`:

            - 0 corresponds to a *sentence A* token,
            - 1 corresponds to a *sentence B* token.

            [What are token type IDs?](../glossary#token-type-ids)
        position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
            config.max_position_embeddings - 1]`.

            [What are position IDs?](../glossary#position-ids)
        head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
            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 (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*):
            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.
        output_attentions (`bool`, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more detail.
        output_hidden_states (`bool`, *optional*):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more detail.
        return_dict (`bool`, *optional*):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
"""


@add_start_docstrings(
    "The bare XLM-RoBERTa Model transformer outputting raw hidden-states without any specific head on top.",
    XLM_ROBERTA_START_DOCSTRING,
)
# Copied from transformers.models.roberta.modeling_roberta.RobertaModel with Roberta->XLMRoberta, ROBERTA->XLM_ROBERTA
class XLMRobertaDecoupledModel(XLMRobertaPreTrainedModel):
    """

    The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
    cross-attention is added between the self-attention layers, following the architecture described in *Attention is
    all you need*_ by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz
    Kaiser and Illia Polosukhin.

    To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
    to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
    `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.

    .. _*Attention is all you need*: https://arxiv.org/abs/1706.03762

    """

    # Copied from transformers.models.clap.modeling_clap.ClapTextModel.__init__ with ClapText->XLMRoberta
    def __init__(self, config, add_pooling_layer=True):
        super().__init__(config)
        self.config = config

        self.embeddings = XLMRobertaDecoupledEmbeddings(config)
        self.encoder = XLMRobertaEncoder(config)

        self.pooler = XLMRobertaPooler(config) if add_pooling_layer else None

        # Initialize weights and apply final processing
        self.post_init()

    def get_input_embeddings(self):
        return self.embeddings.word_embeddings

    def set_input_embeddings(self, value):
        self.embeddings.word_embeddings = value

    def _prune_heads(self, heads_to_prune):
        """
        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
        class PreTrainedModel
        """
        for layer, heads in heads_to_prune.items():
            self.encoder.layer[layer].attention.prune_heads(heads)

    @add_start_docstrings_to_model_forward(XLM_ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=BaseModelOutputWithPoolingAndCrossAttentions,
        config_class=_CONFIG_FOR_DOC,
    )
    # Copied from transformers.models.clap.modeling_clap.ClapTextModel.forward
    def forward(
            self,
            input_ids: Optional[torch.Tensor] = None,
            attention_mask: Optional[torch.Tensor] = None,
            token_lang_ids: Optional[torch.Tensor] = None,
            token_script_ids: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.Tensor] = None,
            head_mask: Optional[torch.Tensor] = None,
            inputs_embeds: Optional[torch.Tensor] = None,
            encoder_hidden_states: Optional[torch.Tensor] = None,
            encoder_attention_mask: Optional[torch.Tensor] = None,
            past_key_values: Optional[List[torch.FloatTensor]] = None,
            use_cache: Optional[bool] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
        r"""
        encoder_hidden_states  (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
            the model is configured as a decoder.
        encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
            the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:

            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.
        past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
            Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.

            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
        use_cache (`bool`, *optional*):
            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
            `past_key_values`).
        """
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if self.config.is_decoder:
            use_cache = use_cache if use_cache is not None else self.config.use_cache
        else:
            use_cache = False

        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        elif input_ids is not None:
            self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
            input_shape = input_ids.size()
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        batch_size, seq_length = input_shape
        device = input_ids.device if input_ids is not None else inputs_embeds.device

        # past_key_values_length
        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0

        if attention_mask is None:
            attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device)

        if self.config.decouple_at_input_embeddings:
            if token_lang_ids is None and self.config.use_lang_embedding is True:
                raise ValueError("token_lang_ids cannot be None if use_lang_embed is True")
            if token_script_ids is None and self.config.use_script_embedding is True:
                raise ValueError("token_script_ids cannot be None if use_script_embedding is True")

        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
        # ourselves in which case we just need to make it broadcastable to all heads.
        extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape)

        # If a 2D or 3D attention mask is provided for the cross-attention
        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
        if self.config.is_decoder and encoder_hidden_states is not None:
            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
            if encoder_attention_mask is None:
                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
            encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
        else:
            encoder_extended_attention_mask = None

        # Prepare head mask if needed
        # 1.0 in head_mask indicate we keep the head
        # attention_probs has shape bsz x n_heads x N x N
        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

        embedding_output = self.embeddings(
            input_ids=input_ids,
            position_ids=position_ids,
            token_lang_ids=token_lang_ids,
            token_script_ids=token_script_ids,
            inputs_embeds=inputs_embeds,
            past_key_values_length=past_key_values_length,
        )
        encoder_outputs = self.encoder(
            embedding_output,
            attention_mask=extended_attention_mask,
            head_mask=head_mask,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_extended_attention_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        sequence_output = encoder_outputs[0]
        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None

        if not return_dict:
            return (sequence_output, pooled_output) + encoder_outputs[1:]

        return BaseModelOutputWithPoolingAndCrossAttentions(
            last_hidden_state=sequence_output,
            pooler_output=pooled_output,
            past_key_values=encoder_outputs.past_key_values,
            hidden_states=encoder_outputs.hidden_states,
            attentions=encoder_outputs.attentions,
            cross_attentions=encoder_outputs.cross_attentions,
        )


@add_start_docstrings(
    """XLM-RoBERTa Model with a `language modeling` head on top.""",
    XLM_ROBERTA_START_DOCSTRING,
)
# Copied from transformers.models.roberta.modeling_roberta.RobertaForMaskedLM with Roberta->XLMRoberta, ROBERTA->XLM_ROBERTA
class XLMRobertaDecoupledForMaskedLM(XLMRobertaPreTrainedModel):
    _tied_weights_keys = ["lm_head.decoder.weight", "lm_head.decoder.bias"]

    def __init__(self, config):
        super().__init__(config)

        if config.is_decoder:
            logger.warning(
                "If you want to use `XLMRobertaForMaskedLM` make sure `config.is_decoder=False` for "
                "bi-directional self-attention."
            )

        self.roberta = XLMRobertaDecoupledModel(config, add_pooling_layer=False)
        self.lm_head = XLMRobertaDecoupledLMHead(config)
        self.config = config
        # Initialize weights and apply final processing
        self.post_init()

    def get_output_embeddings(self):
        return self.lm_head.decoder

    def set_output_embeddings(self, new_embeddings):
        self.lm_head.decoder = new_embeddings

    @add_start_docstrings_to_model_forward(XLM_ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=MaskedLMOutput,
        config_class=_CONFIG_FOR_DOC,
        mask="<mask>",
        expected_output="' Paris'",
        expected_loss=0.1,
    )
    def forward(
            self,
            input_ids: Optional[torch.LongTensor] = None,
            attention_mask: Optional[torch.FloatTensor] = None,
            token_lang_ids: Optional[torch.Tensor] = None,
            token_script_ids: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            head_mask: Optional[torch.FloatTensor] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            encoder_hidden_states: Optional[torch.FloatTensor] = None,
            encoder_attention_mask: Optional[torch.FloatTensor] = None,
            labels: Optional[torch.LongTensor] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], MaskedLMOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
            config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
            loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
            Used to hide legacy arguments that have been deprecated.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.roberta(
            input_ids,
            attention_mask=attention_mask,
            token_lang_ids=token_lang_ids,
            token_script_ids=token_script_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        sequence_output = outputs[0]

        if self.config.use_lang_embedding:
            if token_lang_ids is None:
                raise ValueError("token_lang_ids cannot be None if use_lang_embedding is True")
            token_lang_embeddings = self.roberta.embeddings.lang_type_embeddings(token_lang_ids)
        else:
            token_lang_embeddings = None

        if self.config.use_script_embedding:
            if token_script_ids is None:
                raise ValueError("token_script_ids cannot be None if use_script_embedding is True")
            token_script_embeddings = self.roberta.embeddings.script_type_embeddings(token_script_ids)
        else:
            token_script_embeddings = None

        # select ouputs and labels
        if labels is not None:
            valid_tokens = labels != -100
            valid_tokens = valid_tokens.unsqueeze(-1).expand_as(sequence_output)
            filtered_output = torch.masked_select(sequence_output, valid_tokens).view(-1, sequence_output.size(-1))
            if token_lang_embeddings is not None:
                token_lang_embeddings = \
                    torch.masked_select(token_lang_embeddings, valid_tokens).view(-1, token_lang_embeddings.size(-1))
            if token_script_embeddings is not None:
                token_script_embeddings = \
                    torch.masked_select(token_script_embeddings, valid_tokens).view(-1, token_script_embeddings.size(-1))
            filtered_labels = torch.masked_select(labels, labels != -100)
            sequence_output = filtered_output
            labels = filtered_labels

        prediction_scores = self.lm_head(sequence_output, token_lang_embeddings, token_script_embeddings)
        masked_lm_loss = None
        if labels is not None:
            # move labels to correct device to enable model parallelism
            labels = labels.to(prediction_scores.device)
            loss_fct = CrossEntropyLoss()
            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))

        if not return_dict:
            output = (prediction_scores,) + outputs[2:]
            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

        return MaskedLMOutput(
            loss=masked_lm_loss,
            logits=prediction_scores,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


# Copied from transformers.models.roberta.modeling_roberta.RobertaLMHead
class XLMRobertaDecoupledLMHead(nn.Module):
    """Roberta Head for masked language modeling."""

    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

        self.decoder = nn.Linear(config.hidden_size, config.vocab_size)
        self.bias = nn.Parameter(torch.zeros(config.vocab_size))
        self.decoder.bias = self.bias
        self.config = config

    def forward(self, features, language_embeddings, script_embeddings, **kwargs):
        if language_embeddings is not None:
            features = features + language_embeddings
        if script_embeddings is not None:
            features = features + script_embeddings
        x = self.dense(features)
        x = gelu(x)
        x = self.layer_norm(x)

        # project back to size of vocabulary with bias
        x = self.decoder(x)

        return x

    def _tie_weights(self):
        # To tie those two weights if they get disconnected (on TPU or when the bias is resized)
        # For accelerate compatibility and to not break backward compatibility
        if self.decoder.bias.device.type == "meta":
            self.decoder.bias = self.bias
        else:
            self.bias = self.decoder.bias


@add_start_docstrings(
    """
    XLM-RoBERTa Model transformer with a sequence classification/regression head on top (a linear layer on top of the
    pooled output) e.g. for GLUE tasks.
    """,
    XLM_ROBERTA_START_DOCSTRING,
)
# Copied from transformers.models.roberta.modeling_roberta.RobertaForSequenceClassification with Roberta->XLMRoberta, ROBERTA->XLM_ROBERTA
class XLMRobertaDecoupledForSequenceClassification(XLMRobertaPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config

        self.roberta = XLMRobertaDecoupledModel(config, add_pooling_layer=False)
        self.classifier = XLMRobertaClassificationHead(config)

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(XLM_ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint="cardiffnlp/twitter-roberta-base-emotion",
        output_type=SequenceClassifierOutput,
        config_class=_CONFIG_FOR_DOC,
        expected_output="'optimism'",
        expected_loss=0.08,
    )
    def forward(
            self,
            input_ids: Optional[torch.LongTensor] = None,
            attention_mask: Optional[torch.FloatTensor] = None,
            token_lang_ids: Optional[torch.Tensor] = None,
            token_script_ids: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            head_mask: Optional[torch.FloatTensor] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            labels: Optional[torch.LongTensor] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], SequenceClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.roberta(
            input_ids,
            attention_mask=attention_mask,
            token_lang_ids=token_lang_ids,
            token_script_ids=token_script_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        sequence_output = outputs[0]
        logits = self.classifier(sequence_output)

        loss = None
        if labels is not None:
            # move labels to correct device to enable model parallelism
            labels = labels.to(logits.device)
            if self.config.problem_type is None:
                if self.num_labels == 1:
                    self.config.problem_type = "regression"
                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
                    self.config.problem_type = "single_label_classification"
                else:
                    self.config.problem_type = "multi_label_classification"

            if self.config.problem_type == "regression":
                loss_fct = MSELoss()
                if self.num_labels == 1:
                    loss = loss_fct(logits.squeeze(), labels.squeeze())
                else:
                    loss = loss_fct(logits, labels)
            elif self.config.problem_type == "single_label_classification":
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(logits, labels)

        if not return_dict:
            output = (logits,) + outputs[2:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutput(
            loss=loss,
            logits=logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


@add_start_docstrings(
    """
    XLM-RoBERTa Model with a multiple choice classification head on top (a linear layer on top of the pooled output and
    a softmax) e.g. for RocStories/SWAG tasks.
    """,
    XLM_ROBERTA_START_DOCSTRING,
)
# Copied from transformers.models.roberta.modeling_roberta.RobertaForMultipleChoice with Roberta->XLMRoberta, ROBERTA->XLM_ROBERTA
class XLMRobertaDecoupledForMultipleChoice(XLMRobertaPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.roberta = XLMRobertaDecoupledModel(config)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.classifier = nn.Linear(config.hidden_size, 1)

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(
        XLM_ROBERTA_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length")
    )
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=MultipleChoiceModelOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
            self,
            input_ids: Optional[torch.LongTensor] = None,
            token_lang_ids: Optional[torch.Tensor] = None,
            token_script_ids: Optional[torch.Tensor] = None,
            attention_mask: Optional[torch.FloatTensor] = None,
            labels: Optional[torch.LongTensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            head_mask: Optional[torch.FloatTensor] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], MultipleChoiceModelOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the multiple choice classification loss. Indices should be in `[0, ...,
            num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See
            `input_ids` above)
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]

        flat_input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None
        flat_position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None
        flat_token_lang_ids = token_lang_ids.view(-1, token_lang_ids.size(-1)) if token_lang_ids is not None else None
        flat_token_script_ids = token_script_ids.view(-1, token_script_ids.size(
            -1)) if token_script_ids is not None else None
        flat_attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None
        flat_inputs_embeds = (
            inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1))
            if inputs_embeds is not None
            else None
        )

        outputs = self.roberta(
            flat_input_ids,
            position_ids=flat_position_ids,
            token_lang_ids=flat_token_lang_ids,
            token_script_ids=flat_token_script_ids,
            attention_mask=flat_attention_mask,
            head_mask=head_mask,
            inputs_embeds=flat_inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        pooled_output = outputs[1]

        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)
        reshaped_logits = logits.view(-1, num_choices)

        loss = None
        if labels is not None:
            # move labels to correct device to enable model parallelism
            labels = labels.to(reshaped_logits.device)
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(reshaped_logits, labels)

        if not return_dict:
            output = (reshaped_logits,) + outputs[2:]
            return ((loss,) + output) if loss is not None else output

        return MultipleChoiceModelOutput(
            loss=loss,
            logits=reshaped_logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


@add_start_docstrings(
    """
    XLM-RoBERTa Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g.
    for Named-Entity-Recognition (NER) tasks.
    """,
    XLM_ROBERTA_START_DOCSTRING,
)
# Copied from transformers.models.roberta.modeling_roberta.RobertaForTokenClassification with Roberta->XLMRoberta, ROBERTA->XLM_ROBERTA
class XLMRobertaDecoupledForTokenClassification(XLMRobertaPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.roberta = XLMRobertaDecoupledModel(config, add_pooling_layer=False)
        classifier_dropout = (
            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(XLM_ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint="Jean-Baptiste/roberta-large-ner-english",
        output_type=TokenClassifierOutput,
        config_class=_CONFIG_FOR_DOC,
        expected_output="['O', 'ORG', 'ORG', 'O', 'O', 'O', 'O', 'O', 'LOC', 'O', 'LOC', 'LOC']",
        expected_loss=0.01,
    )
    def forward(
            self,
            input_ids: Optional[torch.LongTensor] = None,
            attention_mask: Optional[torch.FloatTensor] = None,
            token_lang_ids: Optional[torch.Tensor] = None,
            token_script_ids: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            head_mask: Optional[torch.FloatTensor] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            labels: Optional[torch.LongTensor] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.roberta(
            input_ids,
            attention_mask=attention_mask,
            token_lang_ids=token_lang_ids,
            token_script_ids=token_script_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        sequence_output = outputs[0]

        sequence_output = self.dropout(sequence_output)
        logits = self.classifier(sequence_output)

        loss = None
        if labels is not None:
            # move labels to correct device to enable model parallelism
            labels = labels.to(logits.device)
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

        if not return_dict:
            output = (logits,) + outputs[2:]
            return ((loss,) + output) if loss is not None else output

        return TokenClassifierOutput(
            loss=loss,
            logits=logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


# Copied from transformers.models.roberta.modeling_roberta.RobertaClassificationHead with Roberta->XLMRoberta
class XLMRobertaClassificationHead(nn.Module):
    """Head for sentence-level classification tasks."""

    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        classifier_dropout = (
            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(classifier_dropout)
        self.out_proj = nn.Linear(config.hidden_size, config.num_labels)

    def forward(self, features, **kwargs):
        x = features[:, 0, :]  # take <s> token (equiv. to [CLS])
        x = self.dropout(x)
        x = self.dense(x)
        x = torch.tanh(x)
        x = self.dropout(x)
        x = self.out_proj(x)
        return x


@add_start_docstrings(
    """
    XLM-RoBERTa Model with a span classification head on top for extractive question-answering tasks like SQuAD (a
    linear layers on top of the hidden-states output to compute `span start logits` and `span end logits`).
    """,
    XLM_ROBERTA_START_DOCSTRING,
)
# Copied from transformers.models.roberta.modeling_roberta.RobertaForQuestionAnswering with Roberta->XLMRoberta, ROBERTA->XLM_ROBERTA
class XLMRobertaForQuestionAnswering(XLMRobertaPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.roberta = XLMRobertaDecoupledModel(config, add_pooling_layer=False)
        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(XLM_ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint="deepset/roberta-base-squad2",
        output_type=QuestionAnsweringModelOutput,
        config_class=_CONFIG_FOR_DOC,
        expected_output="' puppet'",
        expected_loss=0.86,
    )
    def forward(
            self,
            input_ids: Optional[torch.LongTensor] = None,
            attention_mask: Optional[torch.FloatTensor] = None,
            token_lang_ids: Optional[torch.Tensor] = None,
            token_script_ids: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            head_mask: Optional[torch.FloatTensor] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            start_positions: Optional[torch.LongTensor] = None,
            end_positions: Optional[torch.LongTensor] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], QuestionAnsweringModelOutput]:
        r"""
        start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the start of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the end of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.roberta(
            input_ids,
            attention_mask=attention_mask,
            token_lang_ids=token_lang_ids,
            token_script_ids=token_script_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        sequence_output = outputs[0]

        logits = self.qa_outputs(sequence_output)
        start_logits, end_logits = logits.split(1, dim=-1)
        start_logits = start_logits.squeeze(-1).contiguous()
        end_logits = end_logits.squeeze(-1).contiguous()

        total_loss = None
        if start_positions is not None and end_positions is not None:
            # If we are on multi-GPU, split add a dimension
            if len(start_positions.size()) > 1:
                start_positions = start_positions.squeeze(-1)
            if len(end_positions.size()) > 1:
                end_positions = end_positions.squeeze(-1)
            # sometimes the start/end positions are outside our model inputs, we ignore these terms
            ignored_index = start_logits.size(1)
            start_positions = start_positions.clamp(0, ignored_index)
            end_positions = end_positions.clamp(0, ignored_index)

            loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
            start_loss = loss_fct(start_logits, start_positions)
            end_loss = loss_fct(end_logits, end_positions)
            total_loss = (start_loss + end_loss) / 2

        if not return_dict:
            output = (start_logits, end_logits) + outputs[2:]
            return ((total_loss,) + output) if total_loss is not None else output

        return QuestionAnsweringModelOutput(
            loss=total_loss,
            start_logits=start_logits,
            end_logits=end_logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


# Copied from transformers.models.roberta.modeling_roberta.create_position_ids_from_input_ids
def create_position_ids_from_input_ids(input_ids, padding_idx, past_key_values_length=0):
    """
    Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
    are ignored. This is modified from fairseq's `utils.make_positions`.

    Args:
        x: torch.Tensor x:

    Returns: torch.Tensor
    """
    # The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
    mask = input_ids.ne(padding_idx).int()
    incremental_indices = (torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length) * mask
    return incremental_indices.long() + padding_idx