train_results.json

{
    "achieved_tflops_per_gpu": 4.448157893301111,
    "achieved_tflops_per_gpu_theoretical": 172.78313270398155,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.2023322582244873,
    "mfu_percent": 0.3143574482898312,
    "mfu_percent_theoretical": 12.21082209922131,
    "total_flos": 1.3906329399721984e+16,
    "train_loss": 0.6722828388214112,
    "train_runtime": 195.3945,
    "train_samples_per_second": 2.559,
    "train_steps_per_second": 0.051,
    "valid_targets_mean": 3909.8,
    "valid_targets_min": 569
}