train_results.json

{
    "achieved_tflops_per_gpu": 1.375884678412727,
    "achieved_tflops_per_gpu_theoretical": 108.41294601034137,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.08443116396665573,
    "mfu_percent": 0.44098867897843813,
    "mfu_percent_theoretical": 34.747739105878644,
    "total_flos": 1.4965206994208686e+18,
    "train_loss": 0.19276038985634514,
    "train_runtime": 135959.8594,
    "train_samples_per_second": 0.506,
    "train_steps_per_second": 0.032,
    "valid_targets_mean": 4994.4,
    "valid_targets_min": 1365
}