train_results.json

{
    "achieved_tflops_per_gpu": 4.808821089255394,
    "achieved_tflops_per_gpu_theoretical": 108.41294601034137,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.34229332208633423,
    "mfu_percent": 0.3398460133749395,
    "mfu_percent_theoretical": 7.661692297550626,
    "total_flos": 1.880116230107431e+18,
    "train_loss": 0.3771306716712417,
    "train_runtime": 24435.7738,
    "train_samples_per_second": 2.046,
    "train_steps_per_second": 0.032,
    "valid_targets_mean": 5900.6,
    "valid_targets_min": 645
}