train_results.json

{
    "achieved_tflops_per_gpu": 4.448451668734935,
    "achieved_tflops_per_gpu_theoretical": 247.31703308609124,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.1196058988571167,
    "mfu_percent": 1.4257857912611973,
    "mfu_percent_theoretical": 79.26827983528565,
    "total_flos": 3.0358571973355766e+18,
    "train_loss": 0.05744029661266912,
    "train_runtime": 85306.5691,
    "train_samples_per_second": 1.16,
    "train_steps_per_second": 0.073,
    "valid_targets_mean": 5778.9,
    "valid_targets_min": 4371
}