train_results.json

{
    "achieved_tflops_per_gpu": 2.7067145335487743,
    "achieved_tflops_per_gpu_theoretical": 88.08551863340236,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.17071497440338135,
    "mfu_percent": 0.867536709470761,
    "mfu_percent_theoretical": 28.232538023526395,
    "total_flos": 1.3627384946606735e+18,
    "train_loss": 0.19606392417490914,
    "train_runtime": 62933.2387,
    "train_samples_per_second": 0.419,
    "train_steps_per_second": 0.026,
    "valid_targets_mean": 16957.1,
    "valid_targets_min": 7571
}