train_results.json

{
    "achieved_tflops_per_gpu": 4.0643769076266185,
    "achieved_tflops_per_gpu_theoretical": 159.2315144526889,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.3454465866088867,
    "mfu_percent": 0.28723511714675753,
    "mfu_percent_theoretical": 11.253110562027485,
    "total_flos": 1.0904976209980948e+18,
    "train_loss": 0.3874767948867409,
    "train_runtime": 16769.139,
    "train_samples_per_second": 2.982,
    "train_steps_per_second": 0.047,
    "valid_targets_mean": 6003.4,
    "valid_targets_min": 835
}