evals/mbert/eval_results_hindi.json

{
    "eval_accuracy": 0.3055626516796526,
    "eval_loss": 3.95077246854643,
    "eval_perplexity": 51.97550077126472
}