train_results.json

{
    "achieved_tflops_per_gpu": 5.200032658316402,
    "achieved_tflops_per_gpu_theoretical": 372.6695019105485,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.0740923136472702,
    "mfu_percent": 0.525786921973347,
    "mfu_percent_theoretical": 37.68144609813433,
    "total_flos": 6.743229377418363e+17,
    "train_loss": 0.20770151239345258,
    "train_runtime": 16209.5842,
    "train_samples_per_second": 1.754,
    "train_steps_per_second": 0.11,
    "valid_targets_mean": 4069.1,
    "valid_targets_min": 2087
}