evals/mbert/eval_results_spanish.json

{
    "eval_accuracy": 0.3220253743973355,
    "eval_loss": 4.129566473031045,
    "eval_perplexity": 62.15097297048524
}