train_results.json

{
    "achieved_tflops_per_gpu": 0.0028888970967278762,
    "achieved_tflops_per_gpu_theoretical": 877.4672817712005,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.0974183976650238,
    "mfu_percent": 0.00020416233899136932,
    "mfu_percent_theoretical": 62.01182203330039,
    "total_flos": 1189064928198656.0,
    "train_loss": 0.2180497848695996,
    "train_runtime": 25724.8893,
    "train_samples_per_second": 4.145,
    "train_steps_per_second": 0.259,
    "valid_targets_mean": 1721.1,
    "valid_targets_min": 760
}