train_results.json

{
    "achieved_tflops_per_gpu": 4.002569748940458,
    "achieved_tflops_per_gpu_theoretical": 233.76541483479863,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.11204883456230164,
    "mfu_percent": 1.282874919532198,
    "mfu_percent_theoretical": 74.92481244705085,
    "total_flos": 3.0947556970284974e+18,
    "train_loss": 0.09641505422623421,
    "train_runtime": 96649.0246,
    "train_samples_per_second": 1.11,
    "train_steps_per_second": 0.069,
    "valid_targets_mean": 4614.6,
    "valid_targets_min": 1320
}