train_results.json

{
    "achieved_tflops_per_gpu": 4.491016057964447,
    "achieved_tflops_per_gpu_theoretical": 504.797779860652,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.18683922290802002,
    "mfu_percent": 0.4540966691571736,
    "mfu_percent_theoretical": 51.041231532927405,
    "total_flos": 1.0544849440651346e+18,
    "train_loss": 0.4722494138848214,
    "train_runtime": 29349.8434,
    "train_samples_per_second": 2.381,
    "train_steps_per_second": 0.149,
    "valid_targets_mean": 3776.8,
    "valid_targets_min": 2389
}