mdeberta-v3-base-triplet-critic-xnli

Para cargar el modelo multitarea:

from transformers import DebertaV2PreTrainedModel, DebertaV2Model
from torch import nn
from transformers.models.deberta_v2.modeling_deberta_v2 import *
from transformers.file_utils import ModelOutput

@dataclass
class TXNLIClassifierOutput(ModelOutput):
    loss: Optional[torch.FloatTensor] = None
    logits: torch.FloatTensor = None
    logits_xnli: torch.FloatTensor = None
    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
    attentions: Optional[Tuple[torch.FloatTensor]] = None

class DebertaV2ForTripletClassification(DebertaV2PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        num_labels = getattr(config, "num_labels", 2)
        self.num_labels = num_labels
        self.deberta = DebertaV2Model(config)
        self.pooler = ContextPooler(config)
        output_dim = self.pooler.output_dim
        self.classifier = nn.Linear(output_dim, num_labels)
        drop_out = getattr(config, "cls_dropout", None)
        drop_out = self.config.hidden_dropout_prob if drop_out is None else drop_out
        self.dropout = StableDropout(drop_out)
        self.classifier_xnli = nn.Linear(output_dim, 3)

        self.post_init()

    def get_input_embeddings(self):
        return self.deberta.get_input_embeddings()

    def set_input_embeddings(self, new_embeddings):
        self.deberta.set_input_embeddings(new_embeddings)

    def forward(
        self,
        input_ids=None,
        attention_mask=None,
        token_type_ids=None,
        position_ids=None,
        inputs_embeds=None,
        labels=None,
        output_attentions=None,
        output_hidden_states=None,
        return_dict=None,
    ):
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        outputs = self.deberta(
            input_ids,
            token_type_ids=token_type_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        encoder_layer = outputs[0]
        pooled_output = self.pooler(encoder_layer)
        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)
        logits_xnli = self.classifier_xnli(pooled_output)
        loss = None
        if labels is not None:
            if labels.dtype != torch.bool:
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            else:
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(logits_xnli.view(-1, 3), labels.view(-1).long())
        if not return_dict:
            output = (logits,) + outputs[1:]
            return ((loss,) + output) if loss is not None else output
        return TXNLIClassifierOutput(
            loss=loss, logits=logits, logits_xnli=logits_xnli, hidden_states=outputs.hidden_states, attentions=outputs.attentions
        )