train_results.json

{
    "achieved_tflops_per_gpu": 2.476802423960275,
    "achieved_tflops_per_gpu_theoretical": 481.0824479208898,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.29739144444465637,
    "mfu_percent": 0.1750390405625636,
    "mfu_percent_theoretical": 33.998759570380905,
    "total_flos": 1.3775029894633226e+18,
    "train_loss": 0.31707094306887884,
    "train_runtime": 34760.1149,
    "train_samples_per_second": 4.545,
    "train_steps_per_second": 0.142,
    "valid_targets_mean": 4348.2,
    "valid_targets_min": 876
}