train_results.json

{
    "achieved_tflops_per_gpu": 5.0287771576248375,
    "achieved_tflops_per_gpu_theoretical": 166.00732357833525,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.1633167266845703,
    "mfu_percent": 0.35539061184627824,
    "mfu_percent_theoretical": 11.731966330624399,
    "total_flos": 1.2782147307999068e+18,
    "train_loss": 0.4453052529981059,
    "train_runtime": 15886.2519,
    "train_samples_per_second": 3.115,
    "train_steps_per_second": 0.049,
    "valid_targets_mean": 3996.9,
    "valid_targets_min": 611
}