train_results.json

{
    "achieved_tflops_per_gpu": 3.938212952590612,
    "achieved_tflops_per_gpu_theoretical": 152.45570532704255,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.34793925285339355,
    "mfu_percent": 0.2783189365788419,
    "mfu_percent_theoretical": 10.77425479343057,
    "total_flos": 1.0929687562016522e+18,
    "train_loss": 0.3876832940775877,
    "train_runtime": 17345.5697,
    "train_samples_per_second": 2.883,
    "train_steps_per_second": 0.045,
    "valid_targets_mean": 3301.4,
    "valid_targets_min": 867
}