# coding=utf-8
# Copyright 2022 The HuggingFace Inc. team. 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 LiLT model."""

import math
from typing import Optional, Tuple, Union

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

from ...activations import ACT2FN
from ...modeling_outputs import (
    BaseModelOutput,
    BaseModelOutputWithPooling,
    QuestionAnsweringModelOutput,
    SequenceClassifierOutput,
    TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings
from .configuration_lilt import LiltConfig


logger = logging.get_logger(__name__)

_CONFIG_FOR_DOC = "LiltConfig"


class LiltTextEmbeddings(nn.Module):
    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 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.register_buffer(
            "position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
        )
        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")

        # 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
        )

    def forward(
        self,
        input_ids=None,
        token_type_ids=None,
        position_ids=None,
        inputs_embeds=None,
    ):
        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 = self.create_position_ids_from_input_ids(input_ids, self.padding_idx).to(
                    input_ids.device
                )
            else:
                position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)

        if input_ids is not None:
            input_shape = input_ids.size()
        else:
            input_shape = inputs_embeds.size()[:-1]

        if token_type_ids is None:
            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
        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, position_ids

    def create_position_ids_from_input_ids(self, input_ids, padding_idx):
        """
        Args:
        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`.
            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)) * mask
        return incremental_indices.long() + padding_idx

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


class LiltLayoutEmbeddings(nn.Module):
    def __init__(self, config):
        super().__init__()
        # we divide the hidden_size by 6 here as there are 6 different layout embeddings,
        # namely left_position, upper_position, right_position, lower_position, height, width
        self.x_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.hidden_size // 6)
        self.y_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.hidden_size // 6)
        self.h_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.hidden_size // 6)
        self.w_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.hidden_size // 6)

        self.padding_idx = config.pad_token_id
        self.box_position_embeddings = nn.Embedding(
            config.max_position_embeddings,
            config.hidden_size // config.channel_shrink_ratio,
            padding_idx=self.padding_idx,
        )
        self.box_linear_embeddings = nn.Linear(
            in_features=config.hidden_size, out_features=config.hidden_size // config.channel_shrink_ratio
        )
        self.LayerNorm = nn.LayerNorm(config.hidden_size // config.channel_shrink_ratio, eps=config.layer_norm_eps)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, bbox=None, position_ids=None):
        try:
            left_position_embeddings = self.x_position_embeddings(bbox[:, :, 0])
            upper_position_embeddings = self.y_position_embeddings(bbox[:, :, 1])
            right_position_embeddings = self.x_position_embeddings(bbox[:, :, 2])
            lower_position_embeddings = self.y_position_embeddings(bbox[:, :, 3])
        except IndexError as e:
            raise IndexError("The `bbox` coordinate values should be within 0-1000 range.") from e

        h_position_embeddings = self.h_position_embeddings(bbox[:, :, 3] - bbox[:, :, 1])
        w_position_embeddings = self.w_position_embeddings(bbox[:, :, 2] - bbox[:, :, 0])

        spatial_position_embeddings = torch.cat(
            [
                left_position_embeddings,
                upper_position_embeddings,
                right_position_embeddings,
                lower_position_embeddings,
                h_position_embeddings,
                w_position_embeddings,
            ],
            dim=-1,
        )
        spatial_position_embeddings = self.box_linear_embeddings(spatial_position_embeddings)
        box_position_embeddings = self.box_position_embeddings(position_ids)

        spatial_position_embeddings = spatial_position_embeddings + box_position_embeddings

        spatial_position_embeddings = self.LayerNorm(spatial_position_embeddings)
        spatial_position_embeddings = self.dropout(spatial_position_embeddings)

        return spatial_position_embeddings


class LiltSelfAttention(nn.Module):
    def __init__(self, config, position_embedding_type=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.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.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.layout_query = nn.Linear(
            config.hidden_size // config.channel_shrink_ratio, self.all_head_size // config.channel_shrink_ratio
        )
        self.layout_key = nn.Linear(
            config.hidden_size // config.channel_shrink_ratio, self.all_head_size // config.channel_shrink_ratio
        )
        self.layout_value = nn.Linear(
            config.hidden_size // config.channel_shrink_ratio, self.all_head_size // config.channel_shrink_ratio
        )

        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
        self.position_embedding_type = position_embedding_type or getattr(
            config, "position_embedding_type", "absolute"
        )
        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
            self.max_position_embeddings = config.max_position_embeddings
            self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size)

        self.channel_shrink_ratio = config.channel_shrink_ratio

    def transpose_for_scores(self, x, r=1):
        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size // r)
        x = x.view(*new_x_shape)
        return x.permute(0, 2, 1, 3)

    def forward(
        self,
        hidden_states,
        layout_inputs,
        attention_mask=None,
        head_mask=None,
        output_attentions=False,
    ):
        layout_value_layer = self.transpose_for_scores(self.layout_value(layout_inputs), r=self.channel_shrink_ratio)
        layout_key_layer = self.transpose_for_scores(self.layout_key(layout_inputs), r=self.channel_shrink_ratio)
        layout_query_layer = self.transpose_for_scores(self.layout_query(layout_inputs), r=self.channel_shrink_ratio)

        mixed_query_layer = self.query(hidden_states)

        key_layer = self.transpose_for_scores(self.key(hidden_states))
        value_layer = self.transpose_for_scores(self.value(hidden_states))
        query_layer = self.transpose_for_scores(mixed_query_layer)

        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
        layout_attention_scores = torch.matmul(layout_query_layer, layout_key_layer.transpose(-1, -2))

        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
            seq_length = hidden_states.size()[1]
            position_ids_l = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(-1, 1)
            position_ids_r = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(1, -1)
            distance = position_ids_l - position_ids_r
            positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
            positional_embedding = positional_embedding.to(dtype=query_layer.dtype)  # fp16 compatibility

            if self.position_embedding_type == "relative_key":
                relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
                attention_scores = attention_scores + relative_position_scores
            elif self.position_embedding_type == "relative_key_query":
                relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
                relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
                attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key

        tmp_attention_scores = attention_scores / math.sqrt(self.attention_head_size)
        tmp_layout_attention_scores = layout_attention_scores / math.sqrt(
            self.attention_head_size // self.channel_shrink_ratio
        )
        attention_scores = tmp_attention_scores + tmp_layout_attention_scores
        layout_attention_scores = tmp_layout_attention_scores + tmp_attention_scores

        if attention_mask is not None:
            # Apply the attention mask is (precomputed for all layers in BertModel forward() function)
            layout_attention_scores = layout_attention_scores + attention_mask

        # Normalize the attention scores to probabilities.
        layout_attention_probs = nn.Softmax(dim=-1)(layout_attention_scores)

        # This is actually dropping out entire tokens to attend to, which might
        # seem a bit unusual, but is taken from the original Transformer paper.
        layout_attention_probs = self.dropout(layout_attention_probs)

        # Mask heads if we want to
        if head_mask is not None:
            layout_attention_probs = layout_attention_probs * head_mask

        layout_context_layer = torch.matmul(layout_attention_probs, layout_value_layer)

        layout_context_layer = layout_context_layer.permute(0, 2, 1, 3).contiguous()
        new_context_layer_shape = layout_context_layer.size()[:-2] + (self.all_head_size // self.channel_shrink_ratio,)
        layout_context_layer = layout_context_layer.view(*new_context_layer_shape)

        if attention_mask is not None:
            # Apply the attention mask is (precomputed for all layers in RobertaModel forward() function)
            attention_scores = attention_scores + attention_mask

        # Normalize the attention scores to probabilities.
        attention_probs = nn.Softmax(dim=-1)(attention_scores)

        # This is actually dropping out entire tokens to attend to, which might
        # seem a bit unusual, but is taken from the original Transformer paper.
        attention_probs = self.dropout(attention_probs)

        # Mask heads if we want to
        if head_mask is not None:
            attention_probs = attention_probs * head_mask

        context_layer = torch.matmul(attention_probs, value_layer)

        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
        context_layer = context_layer.view(*new_context_layer_shape)

        outputs = (
            ((context_layer, layout_context_layer), attention_probs)
            if output_attentions
            else ((context_layer, layout_context_layer),)
        )

        return outputs


# Copied from transformers.models.bert.modeling_bert.BertSelfOutput
class LiltSelfOutput(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 LiltAttention(nn.Module):
    def __init__(self, config, position_embedding_type=None):
        super().__init__()
        self.self = LiltSelfAttention(config, position_embedding_type=position_embedding_type)
        self.output = LiltSelfOutput(config)
        self.pruned_heads = set()

        ori_hidden_size = config.hidden_size
        config.hidden_size = config.hidden_size // config.channel_shrink_ratio
        self.layout_output = LiltSelfOutput(config)
        config.hidden_size = ori_hidden_size

    # Copied from transformers.models.bert.modeling_bert.BertAttention.prune_heads
    def prune_heads(self, heads):
        if len(heads) == 0:
            return
        heads, index = find_pruneable_heads_and_indices(
            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
        )

        # Prune linear layers
        self.self.query = prune_linear_layer(self.self.query, index)
        self.self.key = prune_linear_layer(self.self.key, index)
        self.self.value = prune_linear_layer(self.self.value, index)
        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)

        # Update hyper params and store pruned heads
        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
        self.pruned_heads = self.pruned_heads.union(heads)

    def forward(
        self,
        hidden_states: torch.Tensor,
        layout_inputs: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor]:
        self_outputs = self.self(
            hidden_states,
            layout_inputs,
            attention_mask,
            head_mask,
            output_attentions,
        )
        attention_output = self.output(self_outputs[0][0], hidden_states)
        layout_attention_output = self.layout_output(self_outputs[0][1], layout_inputs)
        outputs = ((attention_output, layout_attention_output),) + self_outputs[1:]  # add attentions if we output them
        return outputs


# Copied from transformers.models.bert.modeling_bert.BertIntermediate
class LiltIntermediate(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


# Copied from transformers.models.bert.modeling_bert.BertOutput
class LiltOutput(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 LiltLayer(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.chunk_size_feed_forward = config.chunk_size_feed_forward
        self.seq_len_dim = 1
        self.attention = LiltAttention(config)
        self.intermediate = LiltIntermediate(config)
        self.output = LiltOutput(config)

        ori_hidden_size = config.hidden_size
        ori_intermediate_size = config.intermediate_size
        config.hidden_size = config.hidden_size // config.channel_shrink_ratio
        config.intermediate_size = config.intermediate_size // config.channel_shrink_ratio
        self.layout_intermediate = LiltIntermediate(config)
        self.layout_output = LiltOutput(config)
        config.hidden_size = ori_hidden_size
        config.intermediate_size = ori_intermediate_size

    def forward(
        self,
        hidden_states: torch.Tensor,
        layout_inputs: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor]:
        self_attention_outputs = self.attention(
            hidden_states,
            layout_inputs,
            attention_mask,
            head_mask,
            output_attentions=output_attentions,
        )
        attention_output = self_attention_outputs[0][0]
        layout_attention_output = self_attention_outputs[0][1]

        outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights

        layer_output = apply_chunking_to_forward(
            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
        )
        layout_layer_output = apply_chunking_to_forward(
            self.layout_feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, layout_attention_output
        )
        outputs = ((layer_output, layout_layer_output),) + outputs

        return outputs

    # Copied from transformers.models.bert.modeling_bert.BertLayer.feed_forward_chunk
    def feed_forward_chunk(self, attention_output):
        intermediate_output = self.intermediate(attention_output)
        layer_output = self.output(intermediate_output, attention_output)
        return layer_output

    def layout_feed_forward_chunk(self, attention_output):
        intermediate_output = self.layout_intermediate(attention_output)
        layer_output = self.layout_output(intermediate_output, attention_output)
        return layer_output


class LiltEncoder(nn.Module):
    # Copied from transformers.models.bert.modeling_bert.BertEncoder.__init__ with Bert->Lilt
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.layer = nn.ModuleList([LiltLayer(config) for _ in range(config.num_hidden_layers)])
        self.gradient_checkpointing = False

    def forward(
        self,
        hidden_states: torch.Tensor,
        layout_inputs: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = False,
        output_hidden_states: Optional[bool] = False,
        return_dict: Optional[bool] = True,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutput]:
        all_hidden_states = () if output_hidden_states else None
        all_self_attentions = () if output_attentions else None

        for i, layer_module in enumerate(self.layer):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            layer_head_mask = head_mask[i] if head_mask is not None else None

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    layer_module.__call__,
                    hidden_states,
                    layout_inputs,
                    attention_mask,
                    layer_head_mask,
                    output_attentions,
                )
            else:
                layer_outputs = layer_module(
                    hidden_states,
                    layout_inputs,
                    attention_mask,
                    layer_head_mask,
                    output_attentions,
                )

            hidden_states = layer_outputs[0][0]
            layout_inputs = layer_outputs[0][1]

            if output_attentions:
                all_self_attentions = all_self_attentions + (layer_outputs[1],)

        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        if not return_dict:
            return tuple(
                v
                for v in [
                    hidden_states,
                    all_hidden_states,
                    all_self_attentions,
                ]
                if v is not None
            )
        return BaseModelOutput(
            last_hidden_state=hidden_states,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions,
        )


# Copied from transformers.models.bert.modeling_bert.BertPooler
class LiltPooler(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 LiltPreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """

    config_class = LiltConfig
    base_model_prefix = "lilt"
    supports_gradient_checkpointing = True
    _no_split_modules = []

    def _init_weights(self, module):
        """Initialize the weights"""
        if isinstance(module, nn.Linear):
            # Slightly different from the TF version which uses truncated_normal for initialization
            # cf https://github.com/pytorch/pytorch/pull/5617
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.padding_idx is not None:
                module.weight.data[module.padding_idx].zero_()
        elif isinstance(module, nn.LayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)


LILT_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 ([`LiltConfig`]): 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.
"""

LILT_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)

        bbox (`torch.LongTensor` of shape `({0}, 4)`, *optional*):
            Bounding boxes of each input sequence tokens. Selected in the range `[0,
            config.max_2d_position_embeddings-1]`. Each bounding box should be a normalized version in (x0, y0, x1, y1)
            format, where (x0, y0) corresponds to the position of the upper left corner in the bounding box, and (x1,
            y1) represents the position of the lower right corner. See [Overview](#Overview) for normalization.

        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 LiLT Model transformer outputting raw hidden-states without any specific head on top.",
    LILT_START_DOCSTRING,
)
class LiltModel(LiltPreTrainedModel):
    def __init__(self, config, add_pooling_layer=True):
        super().__init__(config)
        self.config = config

        self.embeddings = LiltTextEmbeddings(config)
        self.layout_embeddings = LiltLayoutEmbeddings(config)
        self.encoder = LiltEncoder(config)

        self.pooler = LiltPooler(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(LILT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @replace_return_docstrings(output_type=BaseModelOutputWithPooling, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.Tensor] = None,
        bbox: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        token_type_ids: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        inputs_embeds: Optional[torch.Tensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPooling]:
        r"""

        Returns:

        Examples:

        ```python
        >>> from transformers import AutoTokenizer, AutoModel
        >>> from datasets import load_dataset

        >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")
        >>> model = AutoModel.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")

        >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True)
        >>> example = dataset[0]
        >>> words = example["tokens"]
        >>> boxes = example["bboxes"]

        >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt")

        >>> outputs = model(**encoding)
        >>> last_hidden_states = outputs.last_hidden_state
        ```"""
        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 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

        if bbox is None:
            bbox = torch.zeros(input_shape + (4,), dtype=torch.long, device=device)

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

        if token_type_ids is None:
            if hasattr(self.embeddings, "token_type_ids"):
                buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(batch_size, seq_length)
                token_type_ids = buffered_token_type_ids_expanded
            else:
                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)

        # 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)

        # 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, position_ids = self.embeddings(
            input_ids=input_ids,
            position_ids=position_ids,
            token_type_ids=token_type_ids,
            inputs_embeds=inputs_embeds,
        )

        layout_embedding_output = self.layout_embeddings(bbox=bbox, position_ids=position_ids)

        encoder_outputs = self.encoder(
            embedding_output,
            layout_embedding_output,
            attention_mask=extended_attention_mask,
            head_mask=head_mask,
            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 BaseModelOutputWithPooling(
            last_hidden_state=sequence_output,
            pooler_output=pooled_output,
            hidden_states=encoder_outputs.hidden_states,
            attentions=encoder_outputs.attentions,
        )


@add_start_docstrings(
    """
    LiLT Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
    output) e.g. for GLUE tasks.
    """,
    LILT_START_DOCSTRING,
)
class LiltForSequenceClassification(LiltPreTrainedModel):
    # Copied from transformers.models.roberta.modeling_roberta.RobertaForSequenceClassification.__init__ with Roberta->Lilt, roberta->lilt
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config

        self.lilt = LiltModel(config, add_pooling_layer=False)
        self.classifier = LiltClassificationHead(config)

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

    @add_start_docstrings_to_model_forward(LILT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @replace_return_docstrings(output_type=SequenceClassifierOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        bbox: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = 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).

        Returns:

        Examples:

        ```python
        >>> from transformers import AutoTokenizer, AutoModelForSequenceClassification
        >>> from datasets import load_dataset

        >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")
        >>> model = AutoModelForSequenceClassification.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")

        >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True)
        >>> example = dataset[0]
        >>> words = example["tokens"]
        >>> boxes = example["bboxes"]

        >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt")

        >>> outputs = model(**encoding)
        >>> predicted_class_idx = outputs.logits.argmax(-1).item()
        >>> predicted_class = model.config.id2label[predicted_class_idx]
        ```"""
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.lilt(
            input_ids,
            bbox=bbox,
            attention_mask=attention_mask,
            token_type_ids=token_type_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(
    """
    Lilt 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.
    """,
    LILT_START_DOCSTRING,
)
class LiltForTokenClassification(LiltPreTrainedModel):
    # Copied from transformers.models.roberta.modeling_roberta.RobertaForTokenClassification.__init__ with Roberta->Lilt, roberta->lilt
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.lilt = LiltModel(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(LILT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @replace_return_docstrings(output_type=TokenClassifierOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        bbox: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = 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]`.

        Returns:

        Examples:

        ```python
        >>> from transformers import AutoTokenizer, AutoModelForTokenClassification
        >>> from datasets import load_dataset

        >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")
        >>> model = AutoModelForTokenClassification.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")

        >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True)
        >>> example = dataset[0]
        >>> words = example["tokens"]
        >>> boxes = example["bboxes"]

        >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt")

        >>> outputs = model(**encoding)
        >>> predicted_class_indices = outputs.logits.argmax(-1)
        ```"""
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.lilt(
            input_ids,
            bbox=bbox,
            attention_mask=attention_mask,
            token_type_ids=token_type_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->Lilt
class LiltClassificationHead(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(
    """
    Lilt 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`).
    """,
    LILT_START_DOCSTRING,
)
class LiltForQuestionAnswering(LiltPreTrainedModel):
    # Copied from transformers.models.roberta.modeling_roberta.RobertaForQuestionAnswering.__init__ with Roberta->Lilt, roberta->lilt
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.lilt = LiltModel(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(LILT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @replace_return_docstrings(output_type=QuestionAnsweringModelOutput, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        bbox: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = 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.

        Returns:

        Examples:

        ```python
        >>> from transformers import AutoTokenizer, AutoModelForQuestionAnswering
        >>> from datasets import load_dataset

        >>> tokenizer = AutoTokenizer.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")
        >>> model = AutoModelForQuestionAnswering.from_pretrained("SCUT-DLVCLab/lilt-roberta-en-base")

        >>> dataset = load_dataset("nielsr/funsd-layoutlmv3", split="train", trust_remote_code=True)
        >>> example = dataset[0]
        >>> words = example["tokens"]
        >>> boxes = example["bboxes"]

        >>> encoding = tokenizer(words, boxes=boxes, return_tensors="pt")

        >>> outputs = model(**encoding)

        >>> answer_start_index = outputs.start_logits.argmax()
        >>> answer_end_index = outputs.end_logits.argmax()

        >>> predict_answer_tokens = encoding.input_ids[0, answer_start_index : answer_end_index + 1]
        >>> predicted_answer = tokenizer.decode(predict_answer_tokens)
        ```"""
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.lilt(
            input_ids,
            bbox=bbox,
            attention_mask=attention_mask,
            token_type_ids=token_type_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,
        )


__all__ = [
    "LiltForQuestionAnswering",
    "LiltForSequenceClassification",
    "LiltForTokenClassification",
    "LiltModel",
    "LiltPreTrainedModel",
]