xlm-roberta-large / src /models /bert_multi_spans.py

Upload 61 files

cc9c7ee over 3 years ago

3.17 kB

import torch.nn as nn
from torch.nn import BCEWithLogitsLoss
# from transformers import BertModel, BertPreTrainedModel
from transformers import ElectraPreTrainedModel, ElectraModel
from src.utils.mapper import configmapper


@configmapper.map("models", "bert_multi_spans")
# class BertForMultiSpans(BertPreTrainedModel):
class BertForMultiSpans(ElectraPreTrainedModel):
    def __init__(self, config):
        super(BertForMultiSpans, self).__init__(config)
        # self.bert = BertModel(config)
        self.bert = ElectraModel(config)
        self.num_labels = config.num_labels

        # TODO check with Google if it's normal there is no dropout on the token classifier of SQuAD in the TF version
        # self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
        self.init_weights()

    def forward(
        self,
        input_ids=None,
        attention_mask=None,
        token_type_ids=None,
        position_ids=None,
        head_mask=None,
        inputs_embeds=None,
        start_positions=None,
        end_positions=None,
        output_attentions=None,
        output_hidden_states=None,
    ):
        outputs = self.bert(
            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=None,
        )

        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)
        end_logits = end_logits.squeeze(-1)  # batch_size
        # print(start_logits.shape, end_logits.shape, start_positions.shape, end_positions.shape)

        total_loss = None
        if (
            start_positions is not None and end_positions is not None
        ):  # [batch_size/seq_length]
            # # 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.clamp_(0, ignored_index)
            # end_positions.clamp_(0, ignored_index)

            # start_positions = start_logits.view()

            loss_fct = BCEWithLogitsLoss()

            start_loss = loss = loss_fct(
                start_logits,
                start_positions.float(),
            )
            end_loss = loss = loss_fct(
                end_logits,
                end_positions.float(),
            )
            total_loss = (start_loss + end_loss) / 2

        output = (start_logits, end_logits) + outputs[2:]
        return ((total_loss,) + output) if total_loss is not None else output

	import torch.nn as nn
	from torch.nn import BCEWithLogitsLoss
	# from transformers import BertModel, BertPreTrainedModel
	from transformers import ElectraPreTrainedModel, ElectraModel
	from src.utils.mapper import configmapper


	@configmapper.map("models", "bert_multi_spans")
	# class BertForMultiSpans(BertPreTrainedModel):
	class BertForMultiSpans(ElectraPreTrainedModel):
	def __init__(self, config):
	super(BertForMultiSpans, self).__init__(config)
	# self.bert = BertModel(config)
	self.bert = ElectraModel(config)
	self.num_labels = config.num_labels

	# TODO check with Google if it's normal there is no dropout on the token classifier of SQuAD in the TF version
	# self.dropout = nn.Dropout(config.hidden_dropout_prob)
	self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
	self.init_weights()

	def forward(
	self,
	input_ids=None,
	attention_mask=None,
	token_type_ids=None,
	position_ids=None,
	head_mask=None,
	inputs_embeds=None,
	start_positions=None,
	end_positions=None,
	output_attentions=None,
	output_hidden_states=None,
	):
	outputs = self.bert(
	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=None,
	)

	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)
	end_logits = end_logits.squeeze(-1) # batch_size
	# print(start_logits.shape, end_logits.shape, start_positions.shape, end_positions.shape)

	total_loss = None
	if (
	start_positions is not None and end_positions is not None
	): # [batch_size/seq_length]
	# # 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.clamp_(0, ignored_index)
	# end_positions.clamp_(0, ignored_index)

	# start_positions = start_logits.view()

	loss_fct = BCEWithLogitsLoss()

	start_loss = loss = loss_fct(
	start_logits,
	start_positions.float(),
	)
	end_loss = loss = loss_fct(
	end_logits,
	end_positions.float(),
	)
	total_loss = (start_loss + end_loss) / 2

	output = (start_logits, end_logits) + outputs[2:]
	return ((total_loss,) + output) if total_loss is not None else output