train_results.json

{
    "achieved_tflops_per_gpu": 3.83644556224094,
    "achieved_tflops_per_gpu_theoretical": 382.833215599018,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4559069871902466,
    "mfu_percent": 0.27112689485801694,
    "mfu_percent_theoretical": 27.055350925725655,
    "total_flos": 4.2611210572110234e+17,
    "train_loss": 0.5067302330284361,
    "train_runtime": 6941.844,
    "train_samples_per_second": 7.201,
    "train_steps_per_second": 0.113,
    "valid_targets_mean": 1217.6,
    "valid_targets_min": 645
}