train_results.json

{
    "achieved_tflops_per_gpu": 5.333613089682677,
    "achieved_tflops_per_gpu_theoretical": 277.80817415149977,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4209003150463104,
    "mfu_percent": 0.3769337872567263,
    "mfu_percent_theoretical": 19.633086512473483,
    "total_flos": 8.136605209670451e+17,
    "train_loss": 0.4667121115763476,
    "train_runtime": 9534.5841,
    "train_samples_per_second": 5.242,
    "train_steps_per_second": 0.082,
    "valid_targets_mean": 2253.5,
    "valid_targets_min": 645
}