train_results.json

{
    "achieved_tflops_per_gpu": 1.985635002835916,
    "achieved_tflops_per_gpu_theoretical": 132.12827795010355,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.15162084996700287,
    "mfu_percent": 0.6364214752679218,
    "mfu_percent_theoretical": 42.348807035289596,
    "total_flos": 2.4439457607529267e+18,
    "train_loss": 0.23659130134595135,
    "train_runtime": 153851.6493,
    "train_samples_per_second": 0.628,
    "train_steps_per_second": 0.039,
    "valid_targets_mean": 5459.5,
    "valid_targets_min": 2092
}