from transformers.models.xlm_roberta import XLMRobertaPreTrainedModel, XLMRobertaModel
from transformers.modeling_outputs import TokenClassifierOutput
import torch
from torch import nn
from torch.nn import CrossEntropyLoss
from typing import Optional, Tuple, Union


class XLMRobertaForReferenceSegmentation(XLMRobertaPreTrainedModel):
    _keys_to_ignore_on_load_unexpected = [r"pooler"]
    _keys_to_ignore_on_load_missing = [r"position_ids"]

    def __init__(self, config):
        super().__init__(config)
        self.num_labels_first = config.num_labels_first
        self.num_labels_second = config.num_labels_second
        self.alpha = config.alpha

        self.roberta = XLMRobertaModel(config, add_pooling_layer=False)
        classifier_dropout = (
            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
        )
        self.dropout = nn.Dropout(classifier_dropout)
        self.classifier_first = nn.Linear(config.hidden_size, self.num_labels_first)
        self.classifier_second = nn.Linear(config.hidden_size, self.num_labels_second)

        self.post_init()

    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_first: Optional[torch.LongTensor] = None,
            labels_second: Optional[torch.LongTensor] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
    ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.roberta(
            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_first = self.dropout(sequence_output)
        logits_first = self.classifier_first(sequence_output_first)

        sequence_output_second = self.dropout(sequence_output)
        logits_second = self.classifier_second(sequence_output_second)

        loss = None
        if labels_first is not None and labels_second is not None:
            loss_fct_first = CrossEntropyLoss()
            loss_fct_second = CrossEntropyLoss()
            loss_first = loss_fct_first(logits_first.view(-1, self.num_labels_first), labels_first.view(-1))
            loss_second = loss_fct_second(logits_second.view(-1, self.num_labels_second), labels_second.view(-1))
            loss = loss_first + (self.alpha * loss_second)

        return TokenClassifierOutput(
            loss=loss,
            logits=[logits_first, logits_second],
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )