modeling.py

import torch
import torch.nn as nn
from torch.nn import CrossEntropyLoss, MSELoss
from transformers.modeling_albert import AlbertEmbeddings, \
    AlbertConfig, \
    load_tf_weights_in_albert, \
    AlbertTransformer, \
    AlbertModel, \
    AlbertForSequenceClassification


class AlbertEmbeddingsV2(AlbertEmbeddings):
    def __init__(self, config):
        super().__init__(config)

        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)
        doc_id_dic = torch.load('data_in/cached_doc_id_dict')
        self.doc_id_embeddings = nn.Embedding(len(doc_id_dic), config.embedding_size, padding_idx=1)
        self.LayerNorm = torch.nn.LayerNorm(config.embedding_size, eps=config.layer_norm_eps)

    def forward(self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, doc_ids=None):
        if input_ids is not None:
            input_shape = input_ids.size()
        else:
            input_shape = inputs_embeds.size()[:-1]

        seq_length = input_shape[1]
        device = input_ids.device if input_ids is not None else inputs_embeds.device
        if position_ids is None:
            position_ids = torch.arange(seq_length, dtype=torch.long, device=device)
            position_ids = position_ids.unsqueeze(0).expand(input_shape)
        if token_type_ids is None:
            token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=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)
        doc_id_embeddings = self.doc_id_embeddings(doc_ids)

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


class AlbertModelV2(AlbertModel):

    config_class = AlbertConfig
    load_tf_weights = load_tf_weights_in_albert
    base_model_prefix = "albert"

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

        self.config = config
        self.embeddings = AlbertEmbeddingsV2(config)
        self.encoder = AlbertTransformer(config)
        self.pooler = nn.Linear(config.hidden_size, config.hidden_size)
        self.pooler_activation = nn.Tanh()

        self.init_weights()

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

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

    def _resize_token_embeddings(self, new_num_tokens):
        old_embeddings = self.embeddings.word_embeddings
        new_embeddings = self._get_resized_embeddings(old_embeddings, new_num_tokens)
        self.embeddings.word_embeddings = new_embeddings
        return self.embeddings.word_embeddings

    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}
            ALBERT has a different architecture in that its layers are shared across groups, which then has inner groups.
            If an ALBERT model has 12 hidden layers and 2 hidden groups, with two inner groups, there
            is a total of 4 different layers.

            These layers are flattened: the indices [0,1] correspond to the two inner groups of the first hidden layer,
            while [2,3] correspond to the two inner groups of the second hidden layer.

            Any layer with in index other than [0,1,2,3] will result in an error.
            See base class PreTrainedModel for more information about head pruning
        """
        for layer, heads in heads_to_prune.items():
            group_idx = int(layer / self.config.inner_group_num)
            inner_group_idx = int(layer - group_idx * self.config.inner_group_num)
            self.encoder.albert_layer_groups[group_idx].albert_layers[inner_group_idx].attention.prune_heads(heads)

    def forward(
        self,
        input_ids=None,
        attention_mask=None,
        token_type_ids=None,
        position_ids=None,
        doc_ids=None,
        head_mask=None,
        inputs_embeds=None,
        output_attentions=None,
        output_hidden_states=None,
    ):
        r"""
    Return:
        :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.AlbertConfig`) and inputs:
        last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
            Sequence of hidden-states at the output of the last layer of the model.
        pooler_output (:obj:`torch.FloatTensor`: of shape :obj:`(batch_size, hidden_size)`):
            Last layer hidden-state of the first token of the sequence (classification token)
            further processed by a Linear layer and a Tanh activation function. The Linear
            layer weights are trained from the next sentence prediction (classification)
            objective during pre-training.

            This output is usually *not* a good summary
            of the semantic content of the input, you're often better with averaging or pooling
            the sequence of hidden-states for the whole input sequence.
        hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
            of shape :obj:`(batch_size, sequence_length, hidden_size)`.

            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.

            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
            heads.
        """

        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
        )

        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:
            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")

        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:
            token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)

        extended_attention_mask = attention_mask.unsqueeze(1).unsqueeze(2)
        extended_attention_mask = extended_attention_mask.to(dtype=self.dtype)  # fp16 compatibility
        extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0
        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

        embedding_output = self.embeddings(
            input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds, doc_ids=doc_ids
        )
        encoder_outputs = self.encoder(
            embedding_output,
            extended_attention_mask,
            head_mask=head_mask,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
        )

        sequence_output = encoder_outputs[0]

        pooled_output = self.pooler_activation(self.pooler(sequence_output[:, 0]))

        outputs = (sequence_output, pooled_output) + encoder_outputs[
            1:
        ]  # add hidden_states and attentions if they are here
        return outputs


class AlbertForSequenceClassificationV2(AlbertForSequenceClassification):
    def __init__(self, config):
        super(AlbertForSequenceClassification, self).__init__(config)
        self.num_labels = config.num_labels

        self.albert = AlbertModelV2(config)
        self.dropout = nn.Dropout(config.classifier_dropout_prob)
        self.classifier = nn.Linear(config.hidden_size, self.config.num_labels)

        self.init_weights()

    def forward(
        self,
        input_ids=None,
        attention_mask=None,
        token_type_ids=None,
        position_ids=None,
        doc_ids=None,
        head_mask=None,
        inputs_embeds=None,
        labels=None,
        output_attentions=None,
        output_hidden_states=None,
    ):
        r"""
        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`, defaults to :obj:`None`):
            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:
        :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.AlbertConfig`) and inputs:
        loss: (`optional`, returned when ``labels`` is provided) ``torch.FloatTensor`` of shape ``(1,)``:
            Classification (or regression if config.num_labels==1) loss.
        logits ``torch.FloatTensor`` of shape ``(batch_size, config.num_labels)``
            Classification (or regression if config.num_labels==1) scores (before SoftMax).
        hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
            of shape :obj:`(batch_size, sequence_length, hidden_size)`.

            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.

            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
            heads.
        """

        outputs = self.albert(
            input_ids=input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            doc_ids=doc_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
        )

        pooled_output = outputs[1]

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

        outputs = (logits,) + outputs[2:]  # add hidden states and attention if they are here

        if labels is not None:
            if self.num_labels == 1:
                #  We are doing regression
                loss_fct = MSELoss()
                loss = loss_fct(logits.view(-1), labels.view(-1))
            else:
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            outputs = (loss,) + outputs

        return outputs  # (loss), logits, (hidden_states), (attentions)