transformers_4_35_0/models/convbert/modeling_convbert.py

# coding=utf-8
# Copyright 2021 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 ConvBERT model."""


import math
import os
from operator import attrgetter
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, get_activation
from ...modeling_outputs import (
    BaseModelOutputWithCrossAttentions,
    MaskedLMOutput,
    MultipleChoiceModelOutput,
    QuestionAnsweringModelOutput,
    SequenceClassifierOutput,
    TokenClassifierOutput,
)
from ...modeling_utils import PreTrainedModel, SequenceSummary
from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
from ...utils import add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward, logging
from .configuration_convbert import ConvBertConfig


logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "YituTech/conv-bert-base"
_CONFIG_FOR_DOC = "ConvBertConfig"

CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
    "YituTech/conv-bert-base",
    "YituTech/conv-bert-medium-small",
    "YituTech/conv-bert-small",
    # See all ConvBERT models at https://huggingface.co/models?filter=convbert
]


def load_tf_weights_in_convbert(model, config, tf_checkpoint_path):
    """Load tf checkpoints in a pytorch model."""
    try:
        import tensorflow as tf
    except ImportError:
        logger.error(
            "Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see "
            "https://www.tensorflow.org/install/ for installation instructions."
        )
        raise
    tf_path = os.path.abspath(tf_checkpoint_path)
    logger.info(f"Converting TensorFlow checkpoint from {tf_path}")
    # Load weights from TF model
    init_vars = tf.train.list_variables(tf_path)
    tf_data = {}
    for name, shape in init_vars:
        logger.info(f"Loading TF weight {name} with shape {shape}")
        array = tf.train.load_variable(tf_path, name)
        tf_data[name] = array

    param_mapping = {
        "embeddings.word_embeddings.weight": "electra/embeddings/word_embeddings",
        "embeddings.position_embeddings.weight": "electra/embeddings/position_embeddings",
        "embeddings.token_type_embeddings.weight": "electra/embeddings/token_type_embeddings",
        "embeddings.LayerNorm.weight": "electra/embeddings/LayerNorm/gamma",
        "embeddings.LayerNorm.bias": "electra/embeddings/LayerNorm/beta",
        "embeddings_project.weight": "electra/embeddings_project/kernel",
        "embeddings_project.bias": "electra/embeddings_project/bias",
    }
    if config.num_groups > 1:
        group_dense_name = "g_dense"
    else:
        group_dense_name = "dense"

    for j in range(config.num_hidden_layers):
        param_mapping[
            f"encoder.layer.{j}.attention.self.query.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/query/kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.query.bias"
        ] = f"electra/encoder/layer_{j}/attention/self/query/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.self.key.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/key/kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.key.bias"
        ] = f"electra/encoder/layer_{j}/attention/self/key/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.self.value.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/value/kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.value.bias"
        ] = f"electra/encoder/layer_{j}/attention/self/value/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.self.key_conv_attn_layer.depthwise.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_key/depthwise_kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.key_conv_attn_layer.pointwise.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_key/pointwise_kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.key_conv_attn_layer.bias"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_key/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.self.conv_kernel_layer.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_kernel/kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.conv_kernel_layer.bias"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_kernel/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.self.conv_out_layer.weight"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_point/kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.self.conv_out_layer.bias"
        ] = f"electra/encoder/layer_{j}/attention/self/conv_attn_point/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.output.dense.weight"
        ] = f"electra/encoder/layer_{j}/attention/output/dense/kernel"
        param_mapping[
            f"encoder.layer.{j}.attention.output.LayerNorm.weight"
        ] = f"electra/encoder/layer_{j}/attention/output/LayerNorm/gamma"
        param_mapping[
            f"encoder.layer.{j}.attention.output.dense.bias"
        ] = f"electra/encoder/layer_{j}/attention/output/dense/bias"
        param_mapping[
            f"encoder.layer.{j}.attention.output.LayerNorm.bias"
        ] = f"electra/encoder/layer_{j}/attention/output/LayerNorm/beta"
        param_mapping[
            f"encoder.layer.{j}.intermediate.dense.weight"
        ] = f"electra/encoder/layer_{j}/intermediate/{group_dense_name}/kernel"
        param_mapping[
            f"encoder.layer.{j}.intermediate.dense.bias"
        ] = f"electra/encoder/layer_{j}/intermediate/{group_dense_name}/bias"
        param_mapping[
            f"encoder.layer.{j}.output.dense.weight"
        ] = f"electra/encoder/layer_{j}/output/{group_dense_name}/kernel"
        param_mapping[
            f"encoder.layer.{j}.output.dense.bias"
        ] = f"electra/encoder/layer_{j}/output/{group_dense_name}/bias"
        param_mapping[
            f"encoder.layer.{j}.output.LayerNorm.weight"
        ] = f"electra/encoder/layer_{j}/output/LayerNorm/gamma"
        param_mapping[f"encoder.layer.{j}.output.LayerNorm.bias"] = f"electra/encoder/layer_{j}/output/LayerNorm/beta"

    for param in model.named_parameters():
        param_name = param[0]
        retriever = attrgetter(param_name)
        result = retriever(model)
        tf_name = param_mapping[param_name]
        value = torch.from_numpy(tf_data[tf_name])
        logger.info(f"TF: {tf_name}, PT: {param_name} ")
        if tf_name.endswith("/kernel"):
            if not tf_name.endswith("/intermediate/g_dense/kernel"):
                if not tf_name.endswith("/output/g_dense/kernel"):
                    value = value.T
        if tf_name.endswith("/depthwise_kernel"):
            value = value.permute(1, 2, 0)  # 2, 0, 1
        if tf_name.endswith("/pointwise_kernel"):
            value = value.permute(2, 1, 0)  # 2, 1, 0
        if tf_name.endswith("/conv_attn_key/bias"):
            value = value.unsqueeze(-1)
        result.data = value
    return model


class ConvBertEmbeddings(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.embedding_size, padding_idx=config.pad_token_id)
        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.embedding_size)
        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.embedding_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.embedding_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: Optional[torch.LongTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
    ) -> torch.LongTensor:
        if input_ids is not None:
            input_shape = input_ids.size()
        else:
            input_shape = inputs_embeds.size()[:-1]

        seq_length = input_shape[1]

        if position_ids is None:
            position_ids = self.position_ids[:, :seq_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"):
                buffered_token_type_ids = self.token_type_ids[:, :seq_length]
                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
                token_type_ids = buffered_token_type_ids_expanded
            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)
        position_embeddings = self.position_embeddings(position_ids)
        token_type_embeddings = self.token_type_embeddings(token_type_ids)

        embeddings = inputs_embeds + position_embeddings + token_type_embeddings
        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings


class ConvBertPreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """

    config_class = ConvBertConfig
    load_tf_weights = load_tf_weights_in_convbert
    base_model_prefix = "convbert"
    supports_gradient_checkpointing = True

    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)

    def _set_gradient_checkpointing(self, module, value=False):
        if isinstance(module, ConvBertEncoder):
            module.gradient_checkpointing = value


class SeparableConv1D(nn.Module):
    """This class implements separable convolution, i.e. a depthwise and a pointwise layer"""

    def __init__(self, config, input_filters, output_filters, kernel_size, **kwargs):
        super().__init__()
        self.depthwise = nn.Conv1d(
            input_filters,
            input_filters,
            kernel_size=kernel_size,
            groups=input_filters,
            padding=kernel_size // 2,
            bias=False,
        )
        self.pointwise = nn.Conv1d(input_filters, output_filters, kernel_size=1, bias=False)
        self.bias = nn.Parameter(torch.zeros(output_filters, 1))

        self.depthwise.weight.data.normal_(mean=0.0, std=config.initializer_range)
        self.pointwise.weight.data.normal_(mean=0.0, std=config.initializer_range)

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        x = self.depthwise(hidden_states)
        x = self.pointwise(x)
        x += self.bias
        return x


class ConvBertSelfAttention(nn.Module):
    def __init__(self, config):
        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})"
            )

        new_num_attention_heads = config.num_attention_heads // config.head_ratio
        if new_num_attention_heads < 1:
            self.head_ratio = config.num_attention_heads
            self.num_attention_heads = 1
        else:
            self.num_attention_heads = new_num_attention_heads
            self.head_ratio = config.head_ratio

        self.conv_kernel_size = config.conv_kernel_size
        if config.hidden_size % self.num_attention_heads != 0:
            raise ValueError("hidden_size should be divisible by num_attention_heads")

        self.attention_head_size = (config.hidden_size // self.num_attention_heads) // 2
        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.key_conv_attn_layer = SeparableConv1D(
            config, config.hidden_size, self.all_head_size, self.conv_kernel_size
        )
        self.conv_kernel_layer = nn.Linear(self.all_head_size, self.num_attention_heads * self.conv_kernel_size)
        self.conv_out_layer = nn.Linear(config.hidden_size, self.all_head_size)

        self.unfold = nn.Unfold(
            kernel_size=[self.conv_kernel_size, 1], padding=[int((self.conv_kernel_size - 1) / 2), 0]
        )

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

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

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
        mixed_query_layer = self.query(hidden_states)
        batch_size = hidden_states.size(0)
        # If this is instantiated as a cross-attention module, the keys
        # and values come from an encoder; the attention mask needs to be
        # such that the encoder's padding tokens are not attended to.
        if encoder_hidden_states is not None:
            mixed_key_layer = self.key(encoder_hidden_states)
            mixed_value_layer = self.value(encoder_hidden_states)
        else:
            mixed_key_layer = self.key(hidden_states)
            mixed_value_layer = self.value(hidden_states)

        mixed_key_conv_attn_layer = self.key_conv_attn_layer(hidden_states.transpose(1, 2))
        mixed_key_conv_attn_layer = mixed_key_conv_attn_layer.transpose(1, 2)

        query_layer = self.transpose_for_scores(mixed_query_layer)
        key_layer = self.transpose_for_scores(mixed_key_layer)
        value_layer = self.transpose_for_scores(mixed_value_layer)
        conv_attn_layer = torch.multiply(mixed_key_conv_attn_layer, mixed_query_layer)

        conv_kernel_layer = self.conv_kernel_layer(conv_attn_layer)
        conv_kernel_layer = torch.reshape(conv_kernel_layer, [-1, self.conv_kernel_size, 1])
        conv_kernel_layer = torch.softmax(conv_kernel_layer, dim=1)

        conv_out_layer = self.conv_out_layer(hidden_states)
        conv_out_layer = torch.reshape(conv_out_layer, [batch_size, -1, self.all_head_size])
        conv_out_layer = conv_out_layer.transpose(1, 2).contiguous().unsqueeze(-1)
        conv_out_layer = nn.functional.unfold(
            conv_out_layer,
            kernel_size=[self.conv_kernel_size, 1],
            dilation=1,
            padding=[(self.conv_kernel_size - 1) // 2, 0],
            stride=1,
        )
        conv_out_layer = conv_out_layer.transpose(1, 2).reshape(
            batch_size, -1, self.all_head_size, self.conv_kernel_size
        )
        conv_out_layer = torch.reshape(conv_out_layer, [-1, self.attention_head_size, self.conv_kernel_size])
        conv_out_layer = torch.matmul(conv_out_layer, conv_kernel_layer)
        conv_out_layer = torch.reshape(conv_out_layer, [-1, self.all_head_size])

        # Take the dot product between "query" and "key" to get the raw attention scores.
        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
        attention_scores = attention_scores / math.sqrt(self.attention_head_size)
        if attention_mask is not None:
            # Apply the attention mask is (precomputed for all layers in ConvBertModel forward() function)
            attention_scores = attention_scores + attention_mask

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

        # 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()

        conv_out = torch.reshape(conv_out_layer, [batch_size, -1, self.num_attention_heads, self.attention_head_size])
        context_layer = torch.cat([context_layer, conv_out], 2)

        # conv and context
        new_context_layer_shape = context_layer.size()[:-2] + (
            self.num_attention_heads * self.attention_head_size * 2,
        )
        context_layer = context_layer.view(*new_context_layer_shape)

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


class ConvBertSelfOutput(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 ConvBertAttention(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.self = ConvBertSelfAttention(config)
        self.output = ConvBertSelfOutput(config)
        self.pruned_heads = set()

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


class GroupedLinearLayer(nn.Module):
    def __init__(self, input_size, output_size, num_groups):
        super().__init__()
        self.input_size = input_size
        self.output_size = output_size
        self.num_groups = num_groups
        self.group_in_dim = self.input_size // self.num_groups
        self.group_out_dim = self.output_size // self.num_groups
        self.weight = nn.Parameter(torch.empty(self.num_groups, self.group_in_dim, self.group_out_dim))
        self.bias = nn.Parameter(torch.empty(output_size))

    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
        batch_size = list(hidden_states.size())[0]
        x = torch.reshape(hidden_states, [-1, self.num_groups, self.group_in_dim])
        x = x.permute(1, 0, 2)
        x = torch.matmul(x, self.weight)
        x = x.permute(1, 0, 2)
        x = torch.reshape(x, [batch_size, -1, self.output_size])
        x = x + self.bias
        return x


class ConvBertIntermediate(nn.Module):
    def __init__(self, config):
        super().__init__()
        if config.num_groups == 1:
            self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
        else:
            self.dense = GroupedLinearLayer(
                input_size=config.hidden_size, output_size=config.intermediate_size, num_groups=config.num_groups
            )
        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 ConvBertOutput(nn.Module):
    def __init__(self, config):
        super().__init__()
        if config.num_groups == 1:
            self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
        else:
            self.dense = GroupedLinearLayer(
                input_size=config.intermediate_size, output_size=config.hidden_size, num_groups=config.num_groups
            )
        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 ConvBertLayer(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 = ConvBertAttention(config)
        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 TypeError(f"{self} should be used as a decoder model if cross attention is added")
            self.crossattention = ConvBertAttention(config)
        self.intermediate = ConvBertIntermediate(config)
        self.output = ConvBertOutput(config)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: Optional[torch.Tensor] = None,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.Tensor, Optional[torch.FloatTensor]]:
        self_attention_outputs = self.attention(
            hidden_states,
            attention_mask,
            head_mask,
            output_attentions=output_attentions,
        )
        attention_output = self_attention_outputs[0]
        outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights

        if self.is_decoder and encoder_hidden_states is not None:
            if not hasattr(self, "crossattention"):
                raise AttributeError(
                    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_outputs = self.crossattention(
                attention_output,
                encoder_attention_mask,
                head_mask,
                encoder_hidden_states,
                output_attentions,
            )
            attention_output = cross_attention_outputs[0]
            outputs = outputs + cross_attention_outputs[1:]  # add cross 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
        )
        outputs = (layer_output,) + outputs
        return outputs

    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 ConvBertEncoder(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.layer = nn.ModuleList([ConvBertLayer(config) for _ in range(config.num_hidden_layers)])
        self.gradient_checkpointing = False

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: Optional[torch.Tensor] = None,
        output_attentions: Optional[bool] = False,
        output_hidden_states: Optional[bool] = False,
        return_dict: Optional[bool] = True,
    ) -> Union[Tuple, BaseModelOutputWithCrossAttentions]:
        all_hidden_states = () if output_hidden_states else None
        all_self_attentions = () if output_attentions else None
        all_cross_attentions = () if output_attentions and self.config.add_cross_attention 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:

                def create_custom_forward(module):
                    def custom_forward(*inputs):
                        return module(*inputs, output_attentions)

                    return custom_forward

                layer_outputs = torch.utils.checkpoint.checkpoint(
                    create_custom_forward(layer_module),
                    hidden_states,
                    attention_mask,
                    layer_head_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                )
            else:
                layer_outputs = layer_module(
                    hidden_states,
                    attention_mask,
                    layer_head_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                    output_attentions,
                )
            hidden_states = layer_outputs[0]
            if output_attentions:
                all_self_attentions = all_self_attentions + (layer_outputs[1],)
                if self.config.add_cross_attention:
                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)

        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, all_cross_attentions]
                if v is not None
            )
        return BaseModelOutputWithCrossAttentions(
            last_hidden_state=hidden_states,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions,
            cross_attentions=all_cross_attentions,
        )


class ConvBertPredictionHeadTransform(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


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

    Parameters:
        config ([`ConvBertConfig`]): 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.
"""

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

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

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


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

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


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

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

            [What are position IDs?](../glossary#position-ids)
        head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
            Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:


            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.

        inputs_embeds (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*):
            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
            is useful if you want more control over how to convert *input_ids* indices into associated vectors than the
            model's internal embedding lookup matrix.
        output_attentions (`bool`, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more detail.
        output_hidden_states (`bool`, *optional*):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more detail.
        return_dict (`bool`, *optional*):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
"""


@add_start_docstrings(
    "The bare ConvBERT Model transformer outputting raw hidden-states without any specific head on top.",
    CONVBERT_START_DOCSTRING,
)
class ConvBertModel(ConvBertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.embeddings = ConvBertEmbeddings(config)

        if config.embedding_size != config.hidden_size:
            self.embeddings_project = nn.Linear(config.embedding_size, config.hidden_size)

        self.encoder = ConvBertEncoder(config)
        self.config = config
        # 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(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=BaseModelOutputWithCrossAttentions,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: 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,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, BaseModelOutputWithCrossAttentions]:
        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 attention_mask is None:
            attention_mask = torch.ones(input_shape, 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)

        extended_attention_mask = self.get_extended_attention_mask(attention_mask, input_shape)
        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

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

        if hasattr(self, "embeddings_project"):
            hidden_states = self.embeddings_project(hidden_states)

        hidden_states = self.encoder(
            hidden_states,
            attention_mask=extended_attention_mask,
            head_mask=head_mask,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        return hidden_states


class ConvBertGeneratorPredictions(nn.Module):
    """Prediction module for the generator, made up of two dense layers."""

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

        self.LayerNorm = nn.LayerNorm(config.embedding_size, eps=config.layer_norm_eps)
        self.dense = nn.Linear(config.hidden_size, config.embedding_size)

    def forward(self, generator_hidden_states: torch.FloatTensor) -> torch.FloatTensor:
        hidden_states = self.dense(generator_hidden_states)
        hidden_states = get_activation("gelu")(hidden_states)
        hidden_states = self.LayerNorm(hidden_states)

        return hidden_states


@add_start_docstrings("""ConvBERT Model with a `language modeling` head on top.""", CONVBERT_START_DOCSTRING)
class ConvBertForMaskedLM(ConvBertPreTrainedModel):
    _tied_weights_keys = ["generator.lm_head.weight"]

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

        self.convbert = ConvBertModel(config)
        self.generator_predictions = ConvBertGeneratorPredictions(config)

        self.generator_lm_head = nn.Linear(config.embedding_size, config.vocab_size)
        # Initialize weights and apply final processing
        self.post_init()

    def get_output_embeddings(self):
        return self.generator_lm_head

    def set_output_embeddings(self, word_embeddings):
        self.generator_lm_head = word_embeddings

    @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=MaskedLMOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: 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, 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]`
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        generator_hidden_states = self.convbert(
            input_ids,
            attention_mask,
            token_type_ids,
            position_ids,
            head_mask,
            inputs_embeds,
            output_attentions,
            output_hidden_states,
            return_dict,
        )
        generator_sequence_output = generator_hidden_states[0]

        prediction_scores = self.generator_predictions(generator_sequence_output)
        prediction_scores = self.generator_lm_head(prediction_scores)

        loss = None
        # Masked language modeling softmax layer
        if labels is not None:
            loss_fct = nn.CrossEntropyLoss()  # -100 index = padding token
            loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))

        if not return_dict:
            output = (prediction_scores,) + generator_hidden_states[1:]
            return ((loss,) + output) if loss is not None else output

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


class ConvBertClassificationHead(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)

        self.config = config

    def forward(self, hidden_states: torch.Tensor, **kwargs) -> torch.Tensor:
        x = hidden_states[:, 0, :]  # take <s> token (equiv. to [CLS])
        x = self.dropout(x)
        x = self.dense(x)
        x = ACT2FN[self.config.hidden_act](x)
        x = self.dropout(x)
        x = self.out_proj(x)
        return x


@add_start_docstrings(
    """
    ConvBERT Model transformer with a sequence classification/regression head on top (a linear layer on top of the
    pooled output) e.g. for GLUE tasks.
    """,
    CONVBERT_START_DOCSTRING,
)
class ConvBertForSequenceClassification(ConvBertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config
        self.convbert = ConvBertModel(config)
        self.classifier = ConvBertClassificationHead(config)

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

    @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=SequenceClassifierOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: 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, SequenceClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.convbert(
            input_ids,
            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:
            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[1:]
            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(
    """
    ConvBERT Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a
    softmax) e.g. for RocStories/SWAG tasks.
    """,
    CONVBERT_START_DOCSTRING,
)
class ConvBertForMultipleChoice(ConvBertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.convbert = ConvBertModel(config)
        self.sequence_summary = SequenceSummary(config)
        self.classifier = nn.Linear(config.hidden_size, 1)

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

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

        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.convbert(
            input_ids,
            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]

        pooled_output = self.sequence_summary(sequence_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)

        if not return_dict:
            output = (reshaped_logits,) + outputs[1:]
            return ((loss,) + output) if loss is not None else output

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


@add_start_docstrings(
    """
    ConvBERT 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.
    """,
    CONVBERT_START_DOCSTRING,
)
class ConvBertForTokenClassification(ConvBertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.convbert = ConvBertModel(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()

    @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=TokenClassifierOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: 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, TokenClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.convbert(
            input_ids,
            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:
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

        if not return_dict:
            output = (logits,) + outputs[1:]
            return ((loss,) + output) if loss is not None else output

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


@add_start_docstrings(
    """
    ConvBERT 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`).
    """,
    CONVBERT_START_DOCSTRING,
)
class ConvBertForQuestionAnswering(ConvBertPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.num_labels = config.num_labels
        self.convbert = ConvBertModel(config)
        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(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=QuestionAnsweringModelOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: 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, QuestionAnsweringModelOutput]:
        r"""
        start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the start of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the end of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.convbert(
            input_ids,
            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[1:]
            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,
        )