train_results.json

{
    "achieved_tflops_per_gpu": 4.838622773087782,
    "achieved_tflops_per_gpu_theoretical": 111.80085057316455,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.3410820960998535,
    "mfu_percent": 0.3419521394408327,
    "mfu_percent_theoretical": 7.901120181849085,
    "total_flos": 1.8466654367420252e+18,
    "train_loss": 0.38092176382708703,
    "train_runtime": 23853.1903,
    "train_samples_per_second": 2.096,
    "train_steps_per_second": 0.033,
    "valid_targets_mean": 5863.9,
    "valid_targets_min": 645
}