train_results.json

{
    "achieved_tflops_per_gpu": 0.002119835162280068,
    "achieved_tflops_per_gpu_theoretical": 772.4422403236824,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.12605062127113342,
    "mfu_percent": 0.0001498116722459412,
    "mfu_percent_theoretical": 54.58955762004823,
    "total_flos": 649913807929344.0,
    "train_loss": 0.21077685282949843,
    "train_runtime": 19161.6847,
    "train_samples_per_second": 3.639,
    "train_steps_per_second": 0.228,
    "valid_targets_mean": 2205.3,
    "valid_targets_min": 939
}