train_results.json

{
    "achieved_tflops_per_gpu": 3.9026639361996787,
    "achieved_tflops_per_gpu_theoretical": 470.9187342324204,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.35643327236175537,
    "mfu_percent": 0.39460707140542756,
    "mfu_percent_theoretical": 47.615645524006105,
    "total_flos": 1.0328415967578685e+18,
    "train_loss": 0.4849937954541771,
    "train_runtime": 33081.3008,
    "train_samples_per_second": 2.216,
    "train_steps_per_second": 0.139,
    "valid_targets_mean": 2517.8,
    "valid_targets_min": 562
}