train_results.json

{
    "achieved_tflops_per_gpu": 4.074660110932522,
    "achieved_tflops_per_gpu_theoretical": 159.2315144526889,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.40154755115509033,
    "mfu_percent": 0.28796184529558455,
    "mfu_percent_theoretical": 11.253110562027485,
    "total_flos": 1.0927339449708708e+18,
    "train_loss": 0.387720340224588,
    "train_runtime": 16761.1211,
    "train_samples_per_second": 2.983,
    "train_steps_per_second": 0.047,
    "valid_targets_mean": 4463.1,
    "valid_targets_min": 880
}