all_results.json

{
    "achieved_tflops_per_gpu": 3.885773586347376,
    "achieved_tflops_per_gpu_theoretical": 216.82589202068274,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.42353275418281555,
    "mfu_percent": 0.2746129743001679,
    "mfu_percent_theoretical": 15.323384595101253,
    "total_flos": 7.570651310488289e+17,
    "train_loss": 0.4416168404232924,
    "train_runtime": 12176.8728,
    "train_samples_per_second": 4.105,
    "train_steps_per_second": 0.064,
    "valid_targets_mean": 1574.1,
    "valid_targets_min": 271
}