train_results.json

{
    "achieved_tflops_per_gpu": 19.652318907560645,
    "achieved_tflops_per_gpu_theoretical": 1050.2504144751822,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.07527446001768112,
    "mfu_percent": 6.298820162679694,
    "mfu_percent_theoretical": 336.61872258819943,
    "total_flos": 4.1481371263550095e+18,
    "train_loss": 0.012115523323082897,
    "train_runtime": 26384.5271,
    "train_samples_per_second": 4.953,
    "train_steps_per_second": 0.31,
    "valid_targets_mean": 4146.9,
    "valid_targets_min": 1674
}