# modeling.py
from transformers import XLMRobertaForSequenceClassification, AutoConfig
from .modeling_custom import CustomModel

def from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs):
    config = AutoConfig.from_pretrained(pretrained_model_name_or_path)
    num_emotion_labels = config.num_emotion_labels
    model = CustomModel.from_pretrained(pretrained_model_name_or_path, num_emotion_labels=num_emotion_labels, *model_args, **kwargs)
    return model