train_results.json

{
    "achieved_tflops_per_gpu": 9.95198119570938,
    "achieved_tflops_per_gpu_theoretical": 88.08551863340236,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.1424449235200882,
    "mfu_percent": 0.7033202258451858,
    "mfu_percent_theoretical": 6.225124991759883,
    "total_flos": 7.594294192772219e+17,
    "train_loss": 0.5147013063430786,
    "train_runtime": 9538.6713,
    "train_samples_per_second": 1.656,
    "train_steps_per_second": 0.026,
    "valid_targets_mean": 4742.4,
    "valid_targets_min": 1340
}