train_results.json

{
    "achieved_tflops_per_gpu": 2.2220609125909565,
    "achieved_tflops_per_gpu_theoretical": 481.0824479208898,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4729985296726227,
    "mfu_percent": 0.15703610689688738,
    "mfu_percent_theoretical": 33.998759570380905,
    "total_flos": 3.9063951723462656e+17,
    "train_loss": 0.5453641192219889,
    "train_runtime": 10987.534,
    "train_samples_per_second": 4.551,
    "train_steps_per_second": 0.142,
    "valid_targets_mean": 1730.2,
    "valid_targets_min": 414
}