train_results.json

{
    "achieved_tflops_per_gpu": 0.002347270324384033,
    "achieved_tflops_per_gpu_theoretical": 416.7122612272497,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4209580421447754,
    "mfu_percent": 0.00016588482857837688,
    "mfu_percent_theoretical": 29.449629768710224,
    "total_flos": 1065319077380096.0,
    "train_loss": 0.5131396555900574,
    "train_runtime": 28365.903,
    "train_samples_per_second": 1.972,
    "train_steps_per_second": 0.123,
    "valid_targets_mean": 4499.8,
    "valid_targets_min": 1641
}