train_results.json

{
    "achieved_tflops_per_gpu": 4.955181724833952,
    "achieved_tflops_per_gpu_theoretical": 155.84360988986572,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.20247647166252136,
    "mfu_percent": 0.35018952118967855,
    "mfu_percent_theoretical": 11.013682677729026,
    "total_flos": 4.268418036190413e+17,
    "train_loss": 0.4559480676651001,
    "train_runtime": 5383.7809,
    "train_samples_per_second": 2.935,
    "train_steps_per_second": 0.046,
    "valid_targets_mean": 3850.3,
    "valid_targets_min": 771
}