train_results.json

{
    "achieved_tflops_per_gpu": 5.036337465807568,
    "achieved_tflops_per_gpu_theoretical": 159.2315144526889,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.36933785676956177,
    "mfu_percent": 0.3559249092443511,
    "mfu_percent_theoretical": 11.253110562027485,
    "total_flos": 1.3498462023419167e+18,
    "train_loss": 0.4199095495187553,
    "train_runtime": 16751.3373,
    "train_samples_per_second": 2.984,
    "train_steps_per_second": 0.047,
    "valid_targets_mean": 4053.2,
    "valid_targets_min": 645
}