all_results.json

{
    "achieved_tflops_per_gpu": 13.109296686310262,
    "achieved_tflops_per_gpu_theoretical": 1107.844792043176,
    "epoch": 6.999592999592999,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.04449460655450821,
    "mfu_percent": 4.201697655868674,
    "mfu_percent_theoretical": 355.0784589881975,
    "total_flos": 2.7561038524606054e+18,
    "train_loss": 0.015238219677667456,
    "train_runtime": 26280.0507,
    "train_samples_per_second": 5.235,
    "train_steps_per_second": 0.327,
    "valid_targets_mean": 4028.4,
    "valid_targets_min": 1273
}