train_results.json

{
    "achieved_tflops_per_gpu": 4.616871406366803,
    "achieved_tflops_per_gpu_theoretical": 162.61941901551205,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.38632673025131226,
    "mfu_percent": 0.3262806647609048,
    "mfu_percent_theoretical": 11.49253844632594,
    "total_flos": 1.2145665807220736e+18,
    "train_loss": 0.4196553195358082,
    "train_runtime": 16441.9592,
    "train_samples_per_second": 3.04,
    "train_steps_per_second": 0.048,
    "valid_targets_mean": 3155.6,
    "valid_targets_min": 917
}