evals/mbert/eval_results_english.json

{
    "eval_accuracy": 0.2975008599745663,
    "eval_loss": 4.490137260330782,
    "eval_perplexity": 89.13367955737988
}