train_results.json

{
    "achieved_tflops_per_gpu": 0.6998917596550104,
    "achieved_tflops_per_gpu_theoretical": 189.7226555180974,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.25550705194473267,
    "mfu_percent": 0.22432428194070844,
    "mfu_percent_theoretical": 60.80854343528763,
    "total_flos": 4.3179169980769894e+17,
    "train_loss": 0.5313117267349349,
    "train_runtime": 77117.5853,
    "train_samples_per_second": 0.9,
    "train_steps_per_second": 0.056,
    "valid_targets_mean": 1777.2,
    "valid_targets_min": 540
}