classicier.py

import evaluate
import datasets
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

class Classicier(evaluate.Measurement):
    
    def _info(self):
        return evaluate.MeasurementInfo(
            description="",
            citation="",
            inputs_description="",
            features=datasets.Features(
                {
                    "texts": datasets.Value("string", id="sequence"),
                }
            ),
            reference_urls=[],
        )
    
    def _download_and_prepare(self, dl_manager, device=None):
        if device is None:
            device = "cuda" if torch.cuda.is_available() else "cpu"
        
        # Load the tokenizer and model from the specified repository
        self.tokenizer = AutoTokenizer.from_pretrained("AbdulmohsenA/classicier")
        self.model = AutoModelForSequenceClassification.from_pretrained("AbdulmohsenA/classicier")
        
        self.model.to(device)
        self.device = device

    def _compute(self, texts, temperature=2):
        device = self.device
        
        inputs = self.tokenizer(
            texts, 
            return_tensors="pt", 
            truncation=True, 
            padding='max_length', 
            max_length=128
        ).to(device)
        
        with torch.no_grad():
            output = self.model(**inputs)
            prediction = torch.softmax(output.logits / temperature, dim=-1)
            classical_prob = prediction[:, 1].detach().cpu().numpy()

        return {"classical_score": classical_prob}