decoder_module.py

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

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os
import copy
import json
import math
import logging
import tarfile
import tempfile
import shutil
import numpy as np

import torch
from torch import nn


from file_utils import cached_path

logger = logging.getLogger(__name__)

PRETRAINED_MODEL_ARCHIVE_MAP = {
    'decoder-bert-base': "[TODO]",
    'decoder-bert-large': "[TODO]",
}

CONFIG_NAME = 'decoder_bert_config.json'
WEIGHTS_NAME = 'decoder_pytorch_model.bin'

def gelu(x):
    """Implementation of the gelu activation function.
        For information: OpenAI GPT's gelu is slightly different (and gives slightly different results):
        0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3))))
    """
    return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0)))


def swish(x):
    return x * torch.sigmoid(x)


ACT2FN = {"gelu": gelu, "relu": torch.nn.functional.relu, "swish": swish}


class BertConfig(object):
    """Configuration class to store the configuration of a `BertModel`.
    """
    def __init__(self,
                 vocab_size_or_config_json_file,
                 decoder_vocab_size=5,
                 hidden_size=768,
                 num_hidden_layers=12,
                 num_attention_heads=12,
                 intermediate_size=3072,
                 hidden_act="gelu",
                 hidden_dropout_prob=0.1,
                 attention_probs_dropout_prob=0.1,
                 type_vocab_size=2,
                 initializer_range=0.02,
                 max_target_embeddings=128,
                 num_decoder_layers=1):
        """Constructs BertConfig.

        Args:
            vocab_size_or_config_json_file: Vocabulary size of `inputs_ids` in `BertModel`.
            hidden_size: Size of the encoder layers and the pooler layer.
            num_hidden_layers: Number of hidden layers in the Transformer encoder.
            num_attention_heads: Number of attention heads for each attention layer in
                the Transformer encoder.
            intermediate_size: The size of the "intermediate" (i.e., feed-forward)
                layer in the Transformer encoder.
            hidden_act: The non-linear activation function (function or string) in the
                encoder and pooler. If string, "gelu", "relu" and "swish" are supported.
            hidden_dropout_prob: The dropout probabilitiy for all fully connected
                layers in the embeddings, encoder, and pooler.
            attention_probs_dropout_prob: The dropout ratio for the attention
                probabilities.
            type_vocab_size: The vocabulary size of the `token_type_ids` passed into
                `BertModel`.
            initializer_range: The sttdev of the truncated_normal_initializer for
                initializing all weight matrices.
            max_target_embeddings: The maximum sequence length that this model might
                ever be used with. Typically set this to something large just in case
                (e.g., 512 or 1024 or 2048).
            num_decoder_layers:
        """
        if isinstance(vocab_size_or_config_json_file, str):
            with open(vocab_size_or_config_json_file, "r", encoding='utf-8') as reader:
                json_config = json.loads(reader.read())
            for key, value in json_config.items():
                self.__dict__[key] = value
        elif isinstance(vocab_size_or_config_json_file, int):
            self.vocab_size = vocab_size_or_config_json_file
            self.decoder_vocab_size = decoder_vocab_size
            self.hidden_size = hidden_size
            self.num_hidden_layers = num_hidden_layers
            self.num_attention_heads = num_attention_heads
            self.hidden_act = hidden_act
            self.intermediate_size = intermediate_size
            self.hidden_dropout_prob = hidden_dropout_prob
            self.attention_probs_dropout_prob = attention_probs_dropout_prob
            self.type_vocab_size = type_vocab_size
            self.initializer_range = initializer_range
            self.max_target_embeddings = max_target_embeddings
            self.num_decoder_layers = num_decoder_layers
        else:
            raise ValueError("First argument must be either a vocabulary size (int)"
                             "or the path to a pretrained model config file (str)")

    @classmethod
    def from_dict(cls, json_object):
        """Constructs a `BertConfig` from a Python dictionary of parameters."""
        config = BertConfig(vocab_size_or_config_json_file=-1)
        for key, value in json_object.items():
            config.__dict__[key] = value
        return config

    @classmethod
    def from_json_file(cls, json_file):
        """Constructs a `BertConfig` from a json file of parameters."""
        with open(json_file, "r", encoding='utf-8') as reader:
            text = reader.read()
        return cls.from_dict(json.loads(text))

    def __repr__(self):
        return str(self.to_json_string())

    def to_dict(self):
        """Serializes this instance to a Python dictionary."""
        output = copy.deepcopy(self.__dict__)
        return output

    def to_json_string(self):
        """Serializes this instance to a JSON string."""
        return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n"

class DecoderBertLayerNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-12):
        """Construct a layernorm module in the TF style (epsilon inside the square root).
        """
        super(DecoderBertLayerNorm, self).__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.bias = nn.Parameter(torch.zeros(hidden_size))
        self.variance_epsilon = eps

    def forward(self, x):
        u = x.mean(-1, keepdim=True)
        s = (x - u).pow(2).mean(-1, keepdim=True)
        x = (x - u) / torch.sqrt(s + self.variance_epsilon)
        return self.weight * x + self.bias

class BertSelfOutput(nn.Module):
    def __init__(self, config):
        super(BertSelfOutput, self).__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.LayerNorm = DecoderBertLayerNorm(config.hidden_size, eps=1e-12)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

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

    def forward(self, hidden_states):
        hidden_states = self.dense(hidden_states)
        hidden_states = self.intermediate_act_fn(hidden_states)
        return hidden_states


class BertOutput(nn.Module):
    def __init__(self, config):
        super(BertOutput, self).__init__()
        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
        self.LayerNorm = DecoderBertLayerNorm(config.hidden_size, eps=1e-12)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, hidden_states, input_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 BertPredictionHeadTransform(nn.Module):
    def __init__(self, config):
        super(BertPredictionHeadTransform, self).__init__()
        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
        self.transform_act_fn = ACT2FN[config.hidden_act] \
            if isinstance(config.hidden_act, str) else config.hidden_act
        self.LayerNorm = DecoderBertLayerNorm(config.hidden_size, eps=1e-12)

    def forward(self, hidden_states):
        hidden_states = self.dense(hidden_states)
        hidden_states = self.transform_act_fn(hidden_states)
        hidden_states = self.LayerNorm(hidden_states)
        return hidden_states


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

        # The output weights are the same as the input embeddings, but there is
        # an output-only bias for each token.
        self.decoder = nn.Linear(bert_model_embedding_weights.size(1),
                                 bert_model_embedding_weights.size(0),
                                 bias=False)
        self.decoder.weight = bert_model_embedding_weights
        self.bias = nn.Parameter(torch.zeros(bert_model_embedding_weights.size(0)))

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


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

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


class PreTrainedDecoderBertModel(nn.Module):
    """ An abstract class to handle weights initialization and
        a simple interface for dowloading and loading pretrained models.
    """
    def __init__(self, config, *inputs, **kwargs):
        super(PreTrainedDecoderBertModel, self).__init__()
        if not isinstance(config, BertConfig):
            raise ValueError(
                "Parameter config in `{}(config)` should be an instance of class `BertConfig`. "
                "To create a model from a Google pretrained model use "
                "`model = {}.from_pretrained(PRETRAINED_MODEL_NAME)`".format(
                    self.__class__.__name__, self.__class__.__name__
                ))
        self.config = config

    def init_bert_weights(self, module):
        """ Initialize the weights.
        """
        if isinstance(module, (nn.Linear, nn.Embedding)):
            # 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)
        elif isinstance(module, DecoderBertLayerNorm):
            if 'beta' in dir(module) and 'gamma' in dir(module):
                module.beta.data.zero_()
                module.gamma.data.fill_(1.0)
            else:
                module.bias.data.zero_()
                module.weight.data.fill_(1.0)
        if isinstance(module, nn.Linear) and module.bias is not None:
            module.bias.data.zero_()

    def resize_token_embeddings(self, new_num_tokens=None):
        model = self.module if hasattr(self, 'module') else self
        base_model = getattr(model, "bert", model)  # get the base model if needed

        old_embeddings = base_model.embeddings.word_embeddings
        # get_resized_embeddings <-----
        if new_num_tokens is None:
            return old_embeddings

        old_num_tokens, old_embedding_dim = old_embeddings.weight.size()
        if old_num_tokens == new_num_tokens:
            return old_embeddings

        # Build new embeddings
        new_embeddings = nn.Embedding(new_num_tokens, old_embedding_dim)
        device = old_embeddings.weight.device
        new_embeddings.to(device)

        # initialize all new embeddings (in particular added tokens)
        self.init_bert_weights(new_embeddings)

        # Copy word embeddings from the previous weights
        num_tokens_to_copy = min(old_num_tokens, new_num_tokens)
        new_embeddings.weight.data[:num_tokens_to_copy, :] = old_embeddings.weight.data[:num_tokens_to_copy, :]
        # ----->

        # Update base model and current model config
        base_model.embeddings.word_embeddings = new_embeddings
        base_model.config.vocab_size = new_num_tokens
        model.config.vocab_size = new_num_tokens
        model.cls = BertOnlyMLMHead(model.config, base_model.embeddings.word_embeddings.weight).to(device)

    @classmethod
    def get_config(cls, pretrained_model_name, cache_dir=None, type_vocab_size=2, state_dict=None):
        '''
        abstract from `from_pretrained`
        '''
        archive_file = os.path.join(os.path.dirname(os.path.abspath(__file__)), pretrained_model_name)
        if os.path.exists(archive_file) is False:
            if pretrained_model_name in PRETRAINED_MODEL_ARCHIVE_MAP:
                archive_file = PRETRAINED_MODEL_ARCHIVE_MAP[pretrained_model_name]
            else:
                archive_file = pretrained_model_name

        # redirect to the cache, if necessary
        try:
            resolved_archive_file = cached_path(archive_file, cache_dir=cache_dir)
        except FileNotFoundError:
            logger.error(
                "Model name '{}' was not found in model name list. "
                "We assumed '{}' was a path or url but couldn't find any file "
                "associated to this path or url.".format(
                    pretrained_model_name,
                    archive_file))
            return None
        if resolved_archive_file == archive_file:
            logger.info("loading archive file {}".format(archive_file))
        else:
            logger.info("loading archive file {} from cache at {}".format(
                archive_file, resolved_archive_file))
        tempdir = None
        if os.path.isdir(resolved_archive_file):
            serialization_dir = resolved_archive_file
        else:
            # Extract archive to temp dir
            tempdir = tempfile.mkdtemp()
            logger.info("extracting archive file {} to temp dir {}".format(
                resolved_archive_file, tempdir))
            with tarfile.open(resolved_archive_file, 'r:gz') as archive:
                archive.extractall(tempdir)
            serialization_dir = tempdir
        # Load config
        config_file = os.path.join(serialization_dir, CONFIG_NAME)
        config = BertConfig.from_json_file(config_file)
        config.type_vocab_size = type_vocab_size

        if state_dict is None:
            weights_path = os.path.join(serialization_dir, WEIGHTS_NAME)
            if os.path.exists(weights_path):
                state_dict = torch.load(weights_path)

        if tempdir:
            # Clean up temp dir
            shutil.rmtree(tempdir)

        return config, state_dict

    @classmethod
    def init_preweight(cls, model, state_dict):
        old_keys = []
        new_keys = []
        for key in state_dict.keys():
            new_key = None
            if 'gamma' in key:
                new_key = key.replace('gamma', 'weight')
            if 'beta' in key:
                new_key = key.replace('beta', 'bias')
            if new_key:
                old_keys.append(key)
                new_keys.append(new_key)
        for old_key, new_key in zip(old_keys, new_keys):
            state_dict[new_key] = state_dict.pop(old_key)

        missing_keys = []
        unexpected_keys = []
        error_msgs = []
        # copy state_dict so _load_from_state_dict can modify it
        metadata = getattr(state_dict, '_metadata', None)
        state_dict = state_dict.copy()
        if metadata is not None:
            state_dict._metadata = metadata

        def load(module, prefix=''):
            local_metadata = {} if metadata is None else metadata.get(prefix[:-1], {})
            module._load_from_state_dict(
                state_dict, prefix, local_metadata, True, missing_keys, unexpected_keys, error_msgs)
            for name, child in module._modules.items():
                if child is not None:
                    load(child, prefix + name + '.')

        load(model, prefix='' if hasattr(model, 'bert') else 'bert.')
        if len(missing_keys) > 0:
            logger.info("Weights of {} not initialized from pretrained model: {}".format(model.__class__.__name__, "\n   " + "\n   ".join(missing_keys)))
        if len(unexpected_keys) > 0:
            logger.info("Weights from pretrained model not used in {}: {}".format(model.__class__.__name__, "\n   " + "\n   ".join(unexpected_keys)))
        if len(error_msgs) > 0:
            logger.error("Weights from pretrained model cause errors in {}: {}".format(model.__class__.__name__, "\n   " + "\n   ".join(error_msgs)))

        return model

    @classmethod
    def from_pretrained(cls, pretrained_model_name, state_dict=None,
                        cache_dir=None, type_vocab_size=2, *inputs, **kwargs):
        """
        Instantiate a PreTrainedDecoderBertModel from a pre-trained model file or a pytorch state dict.
        Download and cache the pre-trained model file if needed.

        Params:
            pretrained_model_name: either:
                - a str with the name of a pre-trained model to load selected in the list of:
                    . `bert-base-uncased`
                    . `bert-large-uncased`
                    . `bert-base-cased`
                    . `bert-large-cased`
                    . `bert-base-multilingual-uncased`
                    . `bert-base-multilingual-cased`
                    . `bert-base-chinese`
                - a path or url to a pretrained model archive containing:
                    . `bert_config.json` a configuration file for the model
                    . `pytorch_model.bin` a PyTorch dump of a BertForPreTraining instance
            cache_dir: an optional path to a folder in which the pre-trained models will be cached.
            state_dict: an optional state dictionnary (collections.OrderedDict object) to use instead of Google pre-trained models
            *inputs, **kwargs: additional input for the specific Bert class
                (ex: num_classes for BertForSequenceClassification)
        """
        config, state_dict = PreTrainedDecoderBertModel.get_config(pretrained_model_name, cache_dir, type_vocab_size, state_dict)
        # Instantiate model.
        model = cls(config, *inputs, **kwargs)
        if state_dict is None:
            return model
        model = PreTrainedDecoderBertModel.init_preweight(model, state_dict)

        return model

# import torch.nn.functional as F
class ScaledDotProductAttention(nn.Module):
    ''' Scaled Dot-Product Attention '''

    def __init__(self, temperature, attn_dropout=0.1):
        super().__init__()
        self.temperature = temperature
        self.dropout = nn.Dropout(attn_dropout)
        self.softmax = nn.Softmax(dim=2)

    def forward(self, q, k, v, mask=None):

        attn = torch.bmm(q, k.transpose(1, 2))
        attn = attn / self.temperature

        if mask is not None:
            attn = attn.masked_fill(mask, -np.inf)

        attn = self.softmax(attn)
        attn = self.dropout(attn)
        output = torch.bmm(attn, v)

        return output, attn

class MultiHeadAttention(nn.Module):
    ''' Multi-Head Attention module '''

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

        if config.hidden_size % config.num_attention_heads != 0:
            raise ValueError(
                "The hidden size (%d) is not a multiple of the number of attention "
                "heads (%d)" % (config.hidden_size, 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.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, q, k, v, attention_mask):
        mixed_query_layer = self.query(q)
        mixed_key_layer = self.key(k)
        mixed_value_layer = self.value(v)

        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)

        # 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)
        # Apply the attention mask is (precomputed for all layers in BertModel 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)

        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)

        return context_layer, attention_scores

class PositionwiseFeedForward(nn.Module):
    ''' A two-feed-forward-layer module '''

    def __init__(self, d_in, d_hid, dropout=0.1):
        super().__init__()
        self.w_1 = nn.Conv1d(d_in, d_hid, 1) # position-wise
        self.w_2 = nn.Conv1d(d_hid, d_in, 1) # position-wise
        self.layer_norm = nn.LayerNorm(d_in)
        self.dropout = nn.Dropout(dropout)

    def forward(self, x):
        residual = x
        output = x.transpose(1, 2)
        output = self.w_2(gelu(self.w_1(output)))
        output = output.transpose(1, 2)
        output = self.dropout(output)
        output = self.layer_norm(output + residual)
        return output

class DecoderAttention(nn.Module):
    def __init__(self, config):
        super(DecoderAttention, self).__init__()
        self.att = MultiHeadAttention(config)
        self.output = BertSelfOutput(config)

    def forward(self, q, k, v, attention_mask):
        att_output, attention_probs = self.att(q, k, v, attention_mask)
        attention_output = self.output(att_output, q)
        return attention_output, attention_probs

class DecoderLayer(nn.Module):
    def __init__(self, config):
        super(DecoderLayer, self).__init__()
        self.slf_attn = DecoderAttention(config)
        self.enc_attn = DecoderAttention(config)
        self.intermediate = BertIntermediate(config)
        self.output = BertOutput(config)

    def forward(self, dec_input, enc_output, slf_attn_mask=None, dec_enc_attn_mask=None):
        slf_output, _ = self.slf_attn(dec_input, dec_input, dec_input, slf_attn_mask)
        dec_output, dec_att_scores = self.enc_attn(slf_output, enc_output, enc_output, dec_enc_attn_mask)
        intermediate_output = self.intermediate(dec_output)
        dec_output = self.output(intermediate_output, dec_output)
        return dec_output, dec_att_scores

class BertDecoderEmbeddings(nn.Module):
    """Construct the embeddings from word, position and token_type embeddings.
    """
    def __init__(self, config, bert_position_embeddings_weight=None, decoder_word_embeddings_weight=None):
        super(BertDecoderEmbeddings, self).__init__()

        # Design for encoder input
        self.decoder_word_embeddings = nn.Embedding(config.decoder_vocab_size, config.hidden_size)
        if decoder_word_embeddings_weight is not None:
            self.decoder_word_embeddings.weight = decoder_word_embeddings_weight

        self.position_embeddings = nn.Embedding(config.max_target_embeddings, config.hidden_size)
        if bert_position_embeddings_weight is not None:
            self.position_embeddings.weight = bert_position_embeddings_weight

        # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
        # any TensorFlow checkpoint file
        self.LayerNorm = DecoderBertLayerNorm(config.hidden_size, eps=1e-12)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

    def forward(self, input_ids, input_decoder=None):
        seq_length = input_ids.size(1)
        position_ids = torch.arange(seq_length, dtype=torch.long, device=input_ids.device)
        position_ids = position_ids.unsqueeze(0).expand_as(input_ids)

        decoder_words_embeddings = self.decoder_word_embeddings(input_ids)
        position_embeddings = self.position_embeddings(position_ids)

        embeddings = decoder_words_embeddings + position_embeddings
        if input_decoder is not None:
            embeddings = embeddings + input_decoder

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

class BertDecoder(nn.Module):
    def __init__(self, config):
        super(BertDecoder, self).__init__()
        layer = DecoderLayer(config)
        self.layer = nn.ModuleList([copy.deepcopy(layer) for _ in range(config.num_decoder_layers)])

    def forward(self, hidden_states, encoder_outs, self_attn_mask, attention_mask, output_all_encoded_layers=False):
        dec_att_scores = None
        all_encoder_layers = []
        all_dec_att_probs = []
        for layer_module in self.layer:
            hidden_states, dec_att_scores = layer_module(hidden_states, encoder_outs, self_attn_mask, attention_mask)
            if output_all_encoded_layers:
                all_encoder_layers.append(hidden_states)
                all_dec_att_probs.append(dec_att_scores)
        if not output_all_encoded_layers:
            all_encoder_layers.append(hidden_states)
            all_dec_att_probs.append(dec_att_scores)
        return all_encoder_layers, all_dec_att_probs

class BertDecoderClassifier(nn.Module):
    def __init__(self, config, embedding_weights):
        super(BertDecoderClassifier, self).__init__()
        self.cls = BertOnlyMLMHead(config, embedding_weights)

    def forward(self, hidden_states):
        cls_scores = self.cls(hidden_states)
        return cls_scores

class DecoderBertModel(nn.Module):

    def init_decoder_bert_weights(self, module):
        """ Initialize the weights.
        """
        if isinstance(module, (nn.Linear, nn.Embedding)):
            # 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)
        elif isinstance(module, DecoderBertLayerNorm):
            if 'beta' in dir(module) and 'gamma' in dir(module):
                module.beta.data.zero_()
                module.gamma.data.fill_(1.0)
            else:
                module.bias.data.zero_()
                module.weight.data.fill_(1.0)
        if isinstance(module, nn.Linear) and module.bias is not None:
            module.bias.data.zero_()

    """
    Transformer decoder consisting of *args.decoder_layers* layers. Each layer
    is a :class:`TransformerDecoderLayer`.
    
    NOTE: This file is adapted from the Transformer decoder. It is not a decoder used as generation.
    """
    def __init__(self, config, bert_position_embeddings_weight=None, decoder_word_embeddings_weight=None):
        super(DecoderBertModel, self).__init__()
        self.config = config
        self.max_target_length = config.max_target_embeddings
        self.embeddings = BertDecoderEmbeddings(config, bert_position_embeddings_weight, decoder_word_embeddings_weight)
        self.decoder = BertDecoder(config)
        self.apply(self.init_decoder_bert_weights)

    def forward(self, input_ids, encoder_outs=None, answer_mask=None, encoder_mask=None, input_decoder=None):
        """
        Args:
            input_ids (LongTensor): previous decoder outputs of shape `(batch, tgt_len)`, for input feeding/teacher forcing
            encoder_outs (Tensor, optional): output from the encoder, used for encoder-side attention

        Returns:
            tuple:
                - the last decoder layer's output of shape `(batch, tgt_len, vocab)`
                - the last decoder layer's attention weights of shape `(batch, tgt_len, src_len)`
        """
        embedding_output = self.embeddings(input_ids, input_decoder=input_decoder)

        extended_encoder_mask = encoder_mask.unsqueeze(1).unsqueeze(2)   # b x 1 x 1 x ls
        extended_encoder_mask = extended_encoder_mask.to(dtype=next(self.parameters()).dtype) # fp16 compatibility
        extended_encoder_mask = (1.0 - extended_encoder_mask) * -10000.0

        extended_answer_mask = answer_mask.unsqueeze(1).unsqueeze(2)
        extended_answer_mask = extended_answer_mask.to(dtype=next(self.parameters()).dtype)  # fp16 compatibility
        self_attn_mask = (1.0 - extended_answer_mask) * -10000.0

        # NOTE: DecoderBertModel is adapted from the Transformer decoder.
        # It is not a decoder used as generation task. It is used as labeling task here.
        decoded_layers, dec_att_scores = self.decoder(embedding_output,
                                      encoder_outs,
                                      self_attn_mask,
                                      extended_encoder_mask,
                                      )
        sequence_output = decoded_layers[-1]

        return sequence_output