all_results.json

{
    "achieved_tflops_per_gpu": 4.226948326866271,
    "achieved_tflops_per_gpu_theoretical": 203.27427376939005,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.23872976005077362,
    "mfu_percent": 0.42739619078526503,
    "mfu_percent_theoretical": 20.55351605352781,
    "total_flos": 4.410167938263286e+17,
    "train_loss": 0.3955344903163421,
    "train_runtime": 13041.82,
    "train_samples_per_second": 3.807,
    "train_steps_per_second": 0.06,
    "valid_targets_mean": 1017.1,
    "valid_targets_min": 327
}