train_results.json

{
    "achieved_tflops_per_gpu": 3.9092633579818163,
    "achieved_tflops_per_gpu_theoretical": 453.97921141830454,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4973108172416687,
    "mfu_percent": 0.39527435368875796,
    "mfu_percent_theoretical": 45.902852519545455,
    "total_flos": 1.0272989958876692e+18,
    "train_loss": 0.48662940967957485,
    "train_runtime": 32848.2281,
    "train_samples_per_second": 2.142,
    "train_steps_per_second": 0.134,
    "valid_targets_mean": 4057.8,
    "valid_targets_min": 587
}