# Copyright 2018 The Google AI Language Team Authors 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 BERT model."""

from collections.abc import Callable
from dataclasses import dataclass

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

from ... import initialization as init
from ...activations import ACT2FN
from ...cache_utils import Cache, DynamicCache, EncoderDecoderCache
from ...generation import GenerationMixin
from ...masking_utils import create_bidirectional_mask, create_causal_mask
from ...modeling_layers import GradientCheckpointingLayer
from ...modeling_outputs import (
    BaseModelOutputWithPastAndCrossAttentions,
    BaseModelOutputWithPoolingAndCrossAttentions,
    CausalLMOutputWithCrossAttentions,
    MaskedLMOutput,
    MultipleChoiceModelOutput,
    NextSentencePredictorOutput,
    QuestionAnsweringModelOutput,
    SequenceClassifierOutput,
    TokenClassifierOutput,
)
from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
from ...processing_utils import Unpack
from ...pytorch_utils import apply_chunking_to_forward
from ...utils import ModelOutput, TransformersKwargs, auto_docstring, logging
from ...utils.generic import can_return_tuple, check_model_inputs
from .configuration_bert import BertConfig


logger = logging.get_logger(__name__)


class BertEmbeddings(nn.Module):
    """Construct the embeddings from word, position and token_type embeddings."""

    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)

        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.register_buffer(
            "position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
        )
        self.register_buffer(
            "token_type_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
        )

    def forward(
        self,
        input_ids: torch.LongTensor | None = None,
        token_type_ids: torch.LongTensor | None = None,
        position_ids: torch.LongTensor | None = None,
        inputs_embeds: torch.FloatTensor | None = None,
        past_key_values_length: int = 0,
    ) -> torch.Tensor:
        if input_ids is not None:
            input_shape = input_ids.size()
        else:
            input_shape = inputs_embeds.size()[:-1]

        batch_size, seq_length = input_shape

        if position_ids is None:
            position_ids = self.position_ids[:, past_key_values_length : seq_length + past_key_values_length]

        # Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
        # when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
        # issue #5664
        if token_type_ids is None:
            if hasattr(self, "token_type_ids"):
                # NOTE: We assume either pos ids to have bsz == 1 (broadcastable) or bsz == effective bsz (input_shape[0])
                buffered_token_type_ids = self.token_type_ids.expand(position_ids.shape[0], -1)
                buffered_token_type_ids = torch.gather(buffered_token_type_ids, dim=1, index=position_ids)
                token_type_ids = buffered_token_type_ids.expand(batch_size, seq_length)
            else:
                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)

        if inputs_embeds is None:
            inputs_embeds = self.word_embeddings(input_ids)
        token_type_embeddings = self.token_type_embeddings(token_type_ids)
        embeddings = inputs_embeds + token_type_embeddings

        position_embeddings = self.position_embeddings(position_ids)
        embeddings = embeddings + position_embeddings

        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings


def eager_attention_forward(
    module: nn.Module,
    query: torch.Tensor,
    key: torch.Tensor,
    value: torch.Tensor,
    attention_mask: torch.Tensor | None,
    scaling: float | None = None,
    dropout: float = 0.0,
    **kwargs: Unpack[TransformersKwargs],
):
    if scaling is None:
        scaling = query.size(-1) ** -0.5

    # Take the dot product between "query" and "key" to get the raw attention scores.
    attn_weights = torch.matmul(query, key.transpose(2, 3)) * scaling

    if attention_mask is not None:
        attention_mask = attention_mask[:, :, :, : key.shape[-2]]
        attn_weights = attn_weights + attention_mask

    attn_weights = nn.functional.softmax(attn_weights, dim=-1)
    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)

    attn_output = torch.matmul(attn_weights, value)
    attn_output = attn_output.transpose(1, 2).contiguous()

    return attn_output, attn_weights


class BertSelfAttention(nn.Module):
    def __init__(self, config, is_causal=False, layer_idx=None):
        super().__init__()
        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
            raise ValueError(
                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
                f"heads ({config.num_attention_heads})"
            )
        self.config = config

        self.num_attention_heads = config.num_attention_heads
        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
        self.all_head_size = self.num_attention_heads * self.attention_head_size
        self.scaling = self.attention_head_size**-0.5

        self.query = nn.Linear(config.hidden_size, self.all_head_size)
        self.key = nn.Linear(config.hidden_size, self.all_head_size)
        self.value = nn.Linear(config.hidden_size, self.all_head_size)

        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)

        self.is_decoder = config.is_decoder
        self.is_causal = is_causal
        self.layer_idx = layer_idx

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: torch.FloatTensor | None = None,
        past_key_values: Cache | None = None,
        cache_position: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor]:
        input_shape = hidden_states.shape[:-1]
        hidden_shape = (*input_shape, -1, self.attention_head_size)

        # get all proj
        query_layer = self.query(hidden_states).view(*hidden_shape).transpose(1, 2)
        key_layer = self.key(hidden_states).view(*hidden_shape).transpose(1, 2)
        value_layer = self.value(hidden_states).view(*hidden_shape).transpose(1, 2)

        if past_key_values is not None:
            # decoder-only bert can have a simple dynamic cache for example
            current_past_key_values = past_key_values
            if isinstance(past_key_values, EncoderDecoderCache):
                current_past_key_values = past_key_values.self_attention_cache

            # save all key/value_layer to cache to be re-used for fast auto-regressive generation
            key_layer, value_layer = current_past_key_values.update(
                key_layer,
                value_layer,
                self.layer_idx,
                {"cache_position": cache_position},
            )

        attention_interface: Callable = ALL_ATTENTION_FUNCTIONS.get_interface(
            self.config._attn_implementation, eager_attention_forward
        )

        attn_output, attn_weights = attention_interface(
            self,
            query_layer,
            key_layer,
            value_layer,
            attention_mask,
            dropout=0.0 if not self.training else self.dropout.p,
            scaling=self.scaling,
            **kwargs,
        )
        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
        return attn_output, attn_weights


class BertCrossAttention(nn.Module):
    def __init__(self, config, is_causal=False, layer_idx=None):
        super().__init__()
        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
            raise ValueError(
                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
                f"heads ({config.num_attention_heads})"
            )
        self.config = config

        self.num_attention_heads = config.num_attention_heads
        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
        self.all_head_size = self.num_attention_heads * self.attention_head_size
        self.scaling = self.attention_head_size**-0.5

        self.query = nn.Linear(config.hidden_size, self.all_head_size)
        self.key = nn.Linear(config.hidden_size, self.all_head_size)
        self.value = nn.Linear(config.hidden_size, self.all_head_size)

        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)

        self.is_causal = is_causal
        self.layer_idx = layer_idx

    def forward(
        self,
        hidden_states: torch.Tensor,
        encoder_hidden_states: torch.FloatTensor | None = None,
        attention_mask: torch.FloatTensor | None = None,
        past_key_values: EncoderDecoderCache | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor]:
        # determine input shapes
        bsz, tgt_len = hidden_states.shape[:-1]
        src_len = encoder_hidden_states.shape[1]

        q_input_shape = (bsz, tgt_len, -1, self.attention_head_size)
        kv_input_shape = (bsz, src_len, -1, self.attention_head_size)

        # get query proj
        query_layer = self.query(hidden_states).view(*q_input_shape).transpose(1, 2)

        is_updated = past_key_values.is_updated.get(self.layer_idx) if past_key_values is not None else False
        if past_key_values is not None and is_updated:
            # reuse k,v, cross_attentions
            key_layer = past_key_values.cross_attention_cache.layers[self.layer_idx].keys
            value_layer = past_key_values.cross_attention_cache.layers[self.layer_idx].values
        else:
            key_layer = self.key(encoder_hidden_states).view(*kv_input_shape).transpose(1, 2)
            value_layer = self.value(encoder_hidden_states).view(*kv_input_shape).transpose(1, 2)

            if past_key_values is not None:
                # save all states to the cache
                key_layer, value_layer = past_key_values.cross_attention_cache.update(
                    key_layer, value_layer, self.layer_idx
                )
                # set flag that curr layer for cross-attn is already updated so we can re-use in subsequent calls
                past_key_values.is_updated[self.layer_idx] = True

        attention_interface: Callable = ALL_ATTENTION_FUNCTIONS.get_interface(
            self.config._attn_implementation, eager_attention_forward
        )

        attn_output, attn_weights = attention_interface(
            self,
            query_layer,
            key_layer,
            value_layer,
            attention_mask,
            dropout=0.0 if not self.training else self.dropout.p,
            scaling=self.scaling,
            **kwargs,
        )
        attn_output = attn_output.reshape(bsz, tgt_len, -1).contiguous()
        return attn_output, attn_weights


class BertSelfOutput(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states


class BertAttention(nn.Module):
    def __init__(self, config, is_causal=False, layer_idx=None, is_cross_attention=False):
        super().__init__()
        self.is_cross_attention = is_cross_attention
        attention_class = BertCrossAttention if is_cross_attention else BertSelfAttention
        self.self = attention_class(config, is_causal=is_causal, layer_idx=layer_idx)
        self.output = BertSelfOutput(config)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: torch.FloatTensor | None = None,
        encoder_hidden_states: torch.FloatTensor | None = None,
        encoder_attention_mask: torch.FloatTensor | None = None,
        past_key_values: Cache | None = None,
        cache_position: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor]:
        attention_mask = attention_mask if not self.is_cross_attention else encoder_attention_mask
        attention_output, attn_weights = self.self(
            hidden_states,
            encoder_hidden_states=encoder_hidden_states,
            attention_mask=attention_mask,
            past_key_values=past_key_values,
            cache_position=cache_position,
            **kwargs,
        )
        attention_output = self.output(attention_output, hidden_states)
        return attention_output, attn_weights


class BertIntermediate(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
        if isinstance(config.hidden_act, str):
            self.intermediate_act_fn = ACT2FN[config.hidden_act]
        else:
            self.intermediate_act_fn = config.hidden_act

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.intermediate_act_fn(hidden_states)
        return hidden_states


class BertOutput(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states


class BertLayer(GradientCheckpointingLayer):
    def __init__(self, config, layer_idx=None):
        super().__init__()
        self.chunk_size_feed_forward = config.chunk_size_feed_forward
        self.seq_len_dim = 1
        self.attention = BertAttention(config, is_causal=config.is_decoder, layer_idx=layer_idx)
        self.is_decoder = config.is_decoder
        self.add_cross_attention = config.add_cross_attention
        if self.add_cross_attention:
            if not self.is_decoder:
                raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
            self.crossattention = BertAttention(
                config,
                is_causal=False,
                layer_idx=layer_idx,
                is_cross_attention=True,
            )
        self.intermediate = BertIntermediate(config)
        self.output = BertOutput(config)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: torch.FloatTensor | None = None,
        encoder_hidden_states: torch.FloatTensor | None = None,
        encoder_attention_mask: torch.FloatTensor | None = None,
        past_key_values: Cache | None = None,
        cache_position: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor]:
        self_attention_output, _ = self.attention(
            hidden_states,
            attention_mask,
            past_key_values=past_key_values,
            cache_position=cache_position,
            **kwargs,
        )
        attention_output = self_attention_output

        if self.is_decoder and encoder_hidden_states is not None:
            if not hasattr(self, "crossattention"):
                raise ValueError(
                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
                    " by setting `config.add_cross_attention=True`"
                )

            cross_attention_output, _ = self.crossattention(
                self_attention_output,
                None,  # attention_mask
                encoder_hidden_states,
                encoder_attention_mask,
                past_key_values=past_key_values,
                **kwargs,
            )
            attention_output = cross_attention_output

        layer_output = apply_chunking_to_forward(
            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
        )
        return layer_output

    def feed_forward_chunk(self, attention_output):
        intermediate_output = self.intermediate(attention_output)
        layer_output = self.output(intermediate_output, attention_output)
        return layer_output


class BertEncoder(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.layer = nn.ModuleList([BertLayer(config, layer_idx=i) for i in range(config.num_hidden_layers)])

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: torch.FloatTensor | None = None,
        encoder_hidden_states: torch.FloatTensor | None = None,
        encoder_attention_mask: torch.FloatTensor | None = None,
        past_key_values: Cache | None = None,
        use_cache: bool | None = None,
        cache_position: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | BaseModelOutputWithPastAndCrossAttentions:
        for i, layer_module in enumerate(self.layer):
            hidden_states = layer_module(
                hidden_states,
                attention_mask,
                encoder_hidden_states,  # as a positional argument for gradient checkpointing
                encoder_attention_mask=encoder_attention_mask,
                past_key_values=past_key_values,
                cache_position=cache_position,
                **kwargs,
            )

        return BaseModelOutputWithPastAndCrossAttentions(
            last_hidden_state=hidden_states,
            past_key_values=past_key_values if use_cache else None,
        )


class BertPooler(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.activation = nn.Tanh()

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        # We "pool" the model by simply taking the hidden state corresponding
        # to the first token.
        first_token_tensor = hidden_states[:, 0]
        pooled_output = self.dense(first_token_tensor)
        pooled_output = self.activation(pooled_output)
        return pooled_output


class BertPredictionHeadTransform(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        if isinstance(config.hidden_act, str):
            self.transform_act_fn = ACT2FN[config.hidden_act]
        else:
            self.transform_act_fn = config.hidden_act
        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        hidden_states = self.dense(hidden_states)
        hidden_states = self.transform_act_fn(hidden_states)
        hidden_states = self.LayerNorm(hidden_states)
        return hidden_states


class BertLMPredictionHead(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.transform = BertPredictionHeadTransform(config)

        # The output weights are the same as the input embeddings, but there is
        # an output-only bias for each token.
        self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=True)
        self.bias = nn.Parameter(torch.zeros(config.vocab_size))

    def forward(self, hidden_states):
        hidden_states = self.transform(hidden_states)
        hidden_states = self.decoder(hidden_states)
        return hidden_states


class BertOnlyMLMHead(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.predictions = BertLMPredictionHead(config)

    def forward(self, sequence_output: torch.Tensor) -> torch.Tensor:
        prediction_scores = self.predictions(sequence_output)
        return prediction_scores


class BertOnlyNSPHead(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.seq_relationship = nn.Linear(config.hidden_size, 2)

    def forward(self, pooled_output):
        seq_relationship_score = self.seq_relationship(pooled_output)
        return seq_relationship_score


class BertPreTrainingHeads(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.predictions = BertLMPredictionHead(config)
        self.seq_relationship = nn.Linear(config.hidden_size, 2)

    def forward(self, sequence_output, pooled_output):
        prediction_scores = self.predictions(sequence_output)
        seq_relationship_score = self.seq_relationship(pooled_output)
        return prediction_scores, seq_relationship_score


@auto_docstring
class BertPreTrainedModel(PreTrainedModel):
    config_class = BertConfig
    base_model_prefix = "bert"
    supports_gradient_checkpointing = True
    _supports_flash_attn = True
    _supports_sdpa = True
    _supports_flex_attn = True
    _supports_attention_backend = True
    _can_record_outputs = {
        "hidden_states": BertLayer,
        "attentions": BertSelfAttention,
        "cross_attentions": BertCrossAttention,
    }

    @torch.no_grad()
    def _init_weights(self, module):
        """Initialize the weights"""
        super()._init_weights(module)
        if isinstance(module, BertLMPredictionHead):
            init.zeros_(module.bias)
        elif isinstance(module, BertEmbeddings):
            init.copy_(module.position_ids, torch.arange(module.position_ids.shape[-1]).expand((1, -1)))
            init.zeros_(module.token_type_ids)


@dataclass
@auto_docstring(
    custom_intro="""
    Output type of [`BertForPreTraining`].
    """
)
class BertForPreTrainingOutput(ModelOutput):
    r"""
    loss (*optional*, returned when `labels` is provided, `torch.FloatTensor` of shape `(1,)`):
        Total loss as the sum of the masked language modeling loss and the next sequence prediction
        (classification) loss.
    prediction_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
        Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
    seq_relationship_logits (`torch.FloatTensor` of shape `(batch_size, 2)`):
        Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation
        before SoftMax).
    """

    loss: torch.FloatTensor | None = None
    prediction_logits: torch.FloatTensor | None = None
    seq_relationship_logits: torch.FloatTensor | None = None
    hidden_states: tuple[torch.FloatTensor] | None = None
    attentions: tuple[torch.FloatTensor] | None = None


@auto_docstring(
    custom_intro="""
    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](https://huggingface.co/papers/1706.03762) 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.
    """
)
class BertModel(BertPreTrainedModel):
    _no_split_modules = ["BertEmbeddings", "BertLayer"]

    def __init__(self, config, add_pooling_layer=True):
        r"""
        add_pooling_layer (bool, *optional*, defaults to `True`):
            Whether to add a pooling layer
        """
        super().__init__(config)
        self.config = config
        self.gradient_checkpointing = False

        self.embeddings = BertEmbeddings(config)
        self.encoder = BertEncoder(config)

        self.pooler = BertPooler(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

    @check_model_inputs
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        encoder_hidden_states: torch.Tensor | None = None,
        encoder_attention_mask: torch.Tensor | None = None,
        past_key_values: Cache | None = None,
        use_cache: bool | None = None,
        cache_position: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | BaseModelOutputWithPoolingAndCrossAttentions:
        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 use_cache and past_key_values is None:
            past_key_values = (
                EncoderDecoderCache(DynamicCache(config=self.config), DynamicCache(config=self.config))
                if encoder_hidden_states is not None or self.config.is_encoder_decoder
                else DynamicCache(config=self.config)
            )

        if (input_ids is None) ^ (inputs_embeds is not None):
            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")

        if input_ids is not None:
            device = input_ids.device
            seq_length = input_ids.shape[1]
        else:
            device = inputs_embeds.device
            seq_length = inputs_embeds.shape[1]

        past_key_values_length = past_key_values.get_seq_length() if past_key_values is not None else 0
        if cache_position is None:
            cache_position = torch.arange(past_key_values_length, past_key_values_length + seq_length, device=device)

        embedding_output = self.embeddings(
            input_ids=input_ids,
            position_ids=position_ids,
            token_type_ids=token_type_ids,
            inputs_embeds=inputs_embeds,
            past_key_values_length=past_key_values_length,
        )

        attention_mask, encoder_attention_mask = self._create_attention_masks(
            attention_mask=attention_mask,
            encoder_attention_mask=encoder_attention_mask,
            embedding_output=embedding_output,
            encoder_hidden_states=encoder_hidden_states,
            cache_position=cache_position,
            past_key_values=past_key_values,
        )

        encoder_outputs = self.encoder(
            embedding_output,
            attention_mask=attention_mask,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            cache_position=cache_position,
            position_ids=position_ids,
            **kwargs,
        )
        sequence_output = encoder_outputs.last_hidden_state
        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None

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

    def _create_attention_masks(
        self,
        attention_mask,
        encoder_attention_mask,
        embedding_output,
        encoder_hidden_states,
        cache_position,
        past_key_values,
    ):
        if self.config.is_decoder:
            attention_mask = create_causal_mask(
                config=self.config,
                input_embeds=embedding_output,
                attention_mask=attention_mask,
                cache_position=cache_position,
                past_key_values=past_key_values,
            )
        else:
            attention_mask = create_bidirectional_mask(
                config=self.config,
                input_embeds=embedding_output,
                attention_mask=attention_mask,
            )

        if encoder_attention_mask is not None:
            encoder_attention_mask = create_bidirectional_mask(
                config=self.config,
                input_embeds=embedding_output,
                attention_mask=encoder_attention_mask,
                encoder_hidden_states=encoder_hidden_states,
            )

        return attention_mask, encoder_attention_mask


@auto_docstring(
    custom_intro="""
    Bert Model with two heads on top as done during the pretraining: a `masked language modeling` head and a `next
    sentence prediction (classification)` head.
    """
)
class BertForPreTraining(BertPreTrainedModel):
    _tied_weights_keys = {
        "cls.predictions.decoder.weight": "bert.embeddings.word_embeddings.weight",
        "cls.predictions.decoder.bias": "cls.predictions.bias",
    }

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

        self.bert = BertModel(config)
        self.cls = BertPreTrainingHeads(config)

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

    def get_output_embeddings(self):
        return self.cls.predictions.decoder

    def set_output_embeddings(self, new_embeddings):
        self.cls.predictions.decoder = new_embeddings
        self.cls.predictions.bias = new_embeddings.bias

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        next_sentence_label: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | BertForPreTrainingOutput:
        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]`
        next_sentence_label (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the next sequence prediction (classification) loss. Input should be a sequence
            pair (see `input_ids` docstring) Indices should be in `[0, 1]`:

            - 0 indicates sequence B is a continuation of sequence A,
            - 1 indicates sequence B is a random sequence.

        Example:

        ```python
        >>> from transformers import AutoTokenizer, BertForPreTraining
        >>> import torch

        >>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
        >>> model = BertForPreTraining.from_pretrained("google-bert/bert-base-uncased")

        >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
        >>> outputs = model(**inputs)

        >>> prediction_logits = outputs.prediction_logits
        >>> seq_relationship_logits = outputs.seq_relationship_logits
        ```
        """
        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            return_dict=True,
            **kwargs,
        )

        sequence_output, pooled_output = outputs[:2]
        prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output)

        total_loss = None
        if labels is not None and next_sentence_label is not None:
            loss_fct = CrossEntropyLoss()
            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
            next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1))
            total_loss = masked_lm_loss + next_sentence_loss

        return BertForPreTrainingOutput(
            loss=total_loss,
            prediction_logits=prediction_scores,
            seq_relationship_logits=seq_relationship_score,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


@auto_docstring(
    custom_intro="""
    Bert Model with a `language modeling` head on top for CLM fine-tuning.
    """
)
class BertLMHeadModel(BertPreTrainedModel, GenerationMixin):
    _tied_weights_keys = {
        "cls.predictions.decoder.weight": "bert.embeddings.word_embeddings.weight",
        "cls.predictions.decoder.bias": "cls.predictions.bias",
    }

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

        if not config.is_decoder:
            logger.warning("If you want to use `BertLMHeadModel` as a standalone, add `is_decoder=True.`")

        self.bert = BertModel(config, add_pooling_layer=False)
        self.cls = BertOnlyMLMHead(config)

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

    def get_output_embeddings(self):
        return self.cls.predictions.decoder

    def set_output_embeddings(self, new_embeddings):
        self.cls.predictions.decoder = new_embeddings
        self.cls.predictions.bias = new_embeddings.bias

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        encoder_hidden_states: torch.Tensor | None = None,
        encoder_attention_mask: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        past_key_values: Cache | None = None,
        use_cache: bool | None = None,
        cache_position: torch.Tensor | None = None,
        logits_to_keep: int | torch.Tensor = 0,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | CausalLMOutputWithCrossAttentions:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the left-to-right language modeling loss (next word prediction). 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 n `[0, ..., config.vocab_size]`
        """
        if labels is not None:
            use_cache = False

        outputs: BaseModelOutputWithPoolingAndCrossAttentions = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            cache_position=cache_position,
            return_dict=True,
            **kwargs,
        )

        hidden_states = outputs.last_hidden_state
        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
        logits = self.cls(hidden_states[:, slice_indices, :])

        loss = None
        if labels is not None:
            loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)

        return CausalLMOutputWithCrossAttentions(
            loss=loss,
            logits=logits,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
            cross_attentions=outputs.cross_attentions,
        )


@auto_docstring
class BertForMaskedLM(BertPreTrainedModel):
    _tied_weights_keys = {
        "cls.predictions.decoder.weight": "bert.embeddings.word_embeddings.weight",
        "cls.predictions.decoder.bias": "cls.predictions.bias",
    }

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

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

        self.bert = BertModel(config, add_pooling_layer=False)
        self.cls = BertOnlyMLMHead(config)

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

    def get_output_embeddings(self):
        return self.cls.predictions.decoder

    def set_output_embeddings(self, new_embeddings):
        self.cls.predictions.decoder = new_embeddings
        self.cls.predictions.bias = new_embeddings.bias

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        encoder_hidden_states: torch.Tensor | None = None,
        encoder_attention_mask: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> 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]`
        """
        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            return_dict=True,
            **kwargs,
        )

        sequence_output = outputs[0]
        prediction_scores = self.cls(sequence_output)

        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()  # -100 index = padding token
            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))

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


@auto_docstring(
    custom_intro="""
    Bert Model with a `next sentence prediction (classification)` head on top.
    """
)
class BertForNextSentencePrediction(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.bert = BertModel(config)
        self.cls = BertOnlyNSPHead(config)

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

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | NextSentencePredictorOutput:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair
            (see `input_ids` docstring). Indices should be in `[0, 1]`:

            - 0 indicates sequence B is a continuation of sequence A,
            - 1 indicates sequence B is a random sequence.

        Example:

        ```python
        >>> from transformers import AutoTokenizer, BertForNextSentencePrediction
        >>> import torch

        >>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
        >>> model = BertForNextSentencePrediction.from_pretrained("google-bert/bert-base-uncased")

        >>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
        >>> next_sentence = "The sky is blue due to the shorter wavelength of blue light."
        >>> encoding = tokenizer(prompt, next_sentence, return_tensors="pt")

        >>> outputs = model(**encoding, labels=torch.LongTensor([1]))
        >>> logits = outputs.logits
        >>> assert logits[0, 0] < logits[0, 1]  # next sentence was random
        ```
        """

        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            return_dict=True,
            **kwargs,
        )

        pooled_output = outputs[1]

        seq_relationship_scores = self.cls(pooled_output)

        next_sentence_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1))

        return NextSentencePredictorOutput(
            loss=next_sentence_loss,
            logits=seq_relationship_scores,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


@auto_docstring(
    custom_intro="""
    Bert Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
    output) e.g. for GLUE tasks.
    """
)
class BertForSequenceClassification(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config

        self.bert = BertModel(config)
        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()

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> 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).
        """
        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            return_dict=True,
            **kwargs,
        )

        pooled_output = outputs[1]

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

        loss = None
        if labels is not None:
            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)

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


@auto_docstring
class BertForMultipleChoice(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.bert = BertModel(config)
        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, 1)

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

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | MultipleChoiceModelOutput:
        r"""
        input_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`):
            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)
        token_type_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`, *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 `(batch_size, num_choices, sequence_length)`, *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)
        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, num_choices, sequence_length, 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.
        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)
        """
        num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]

        input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None
        attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None
        token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None
        position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None
        inputs_embeds = (
            inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1))
            if inputs_embeds is not None
            else None
        )

        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            return_dict=True,
            **kwargs,
        )

        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:
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(reshaped_logits, labels)

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


@auto_docstring
class BertForTokenClassification(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.bert = BertModel(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()

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        labels: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> 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]`.
        """
        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            return_dict=True,
            **kwargs,
        )

        sequence_output = outputs[0]

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

        loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

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


@auto_docstring
class BertForQuestionAnswering(BertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.bert = BertModel(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()

    @can_return_tuple
    @auto_docstring
    def forward(
        self,
        input_ids: torch.Tensor | None = None,
        attention_mask: torch.Tensor | None = None,
        token_type_ids: torch.Tensor | None = None,
        position_ids: torch.Tensor | None = None,
        inputs_embeds: torch.Tensor | None = None,
        start_positions: torch.Tensor | None = None,
        end_positions: torch.Tensor | None = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> tuple[torch.Tensor] | QuestionAnsweringModelOutput:
        outputs = self.bert(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            return_dict=True,
            **kwargs,
        )

        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

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


__all__ = [
    "BertForMaskedLM",
    "BertForMultipleChoice",
    "BertForNextSentencePrediction",
    "BertForPreTraining",
    "BertForQuestionAnswering",
    "BertForSequenceClassification",
    "BertForTokenClassification",
    "BertLayer",
    "BertLMHeadModel",
    "BertModel",
    "BertPreTrainedModel",
]