train_results.json

{
    "achieved_tflops_per_gpu": 0.00229764142698283,
    "achieved_tflops_per_gpu_theoretical": 779.2180494493286,
    "epoch": 6.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.28063035011291504,
    "mfu_percent": 0.00016237748600585373,
    "mfu_percent_theoretical": 55.06841338864513,
    "total_flos": 598298883194880.0,
    "train_loss": 0.39238279145558674,
    "train_runtime": 16274.8111,
    "train_samples_per_second": 3.685,
    "train_steps_per_second": 0.23,
    "valid_targets_mean": 1914.4,
    "valid_targets_min": 304
}