train_results.json

{
    "achieved_tflops_per_gpu": 8.297900738794809,
    "achieved_tflops_per_gpu_theoretical": 64.37018669364019,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.18857082724571228,
    "mfu_percent": 0.8390192860257643,
    "mfu_percent_theoretical": 6.508613416950475,
    "total_flos": 2.7082456355590636e+18,
    "train_loss": 0.37679352122507276,
    "train_runtime": 40797.1504,
    "train_samples_per_second": 1.226,
    "train_steps_per_second": 0.019,
    "valid_targets_mean": 5942.6,
    "valid_targets_min": 4788
}