train_results.json

{
    "achieved_tflops_per_gpu": 0.0018144536996275832,
    "achieved_tflops_per_gpu_theoretical": 945.225373027664,
    "epoch": 6.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.12043584883213043,
    "mfu_percent": 0.0001282299434365783,
    "mfu_percent_theoretical": 66.80037971926954,
    "total_flos": 254043172110336.0,
    "train_loss": 0.18853386797248878,
    "train_runtime": 8750.677,
    "train_samples_per_second": 4.459,
    "train_steps_per_second": 0.279,
    "valid_targets_mean": 1261.1,
    "valid_targets_min": 857
}