evals/alberti/eval_results_spanish.json

{
    "eval_accuracy": 0.5859480165147395,
    "eval_loss": 2.084652532470857,
    "eval_perplexity": 8.04179673294292
}