train_results.json

{
    "achieved_tflops_per_gpu": 6.582846340644015,
    "achieved_tflops_per_gpu_theoretical": 135.51618251292672,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.3657087981700897,
    "mfu_percent": 0.4652188226603544,
    "mfu_percent_theoretical": 9.577115371938284,
    "total_flos": 2.0706726654792172e+18,
    "train_loss": 0.47480044941993277,
    "train_runtime": 19659.7391,
    "train_samples_per_second": 2.543,
    "train_steps_per_second": 0.04,
    "valid_targets_mean": 6505.4,
    "valid_targets_min": 1469
}