train_results.json

{
    "achieved_tflops_per_gpu": 0.0024687245634862423,
    "achieved_tflops_per_gpu_theoretical": 498.0219707350057,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.45888444781303406,
    "mfu_percent": 0.00017446816703082984,
    "mfu_percent_theoretical": 35.1958989918732,
    "total_flos": 790164860633088.0,
    "train_loss": 0.493048559660664,
    "train_runtime": 20004.3798,
    "train_samples_per_second": 2.354,
    "train_steps_per_second": 0.147,
    "valid_targets_mean": 10751.1,
    "valid_targets_min": 2295
}