train_results.json

{
    "achieved_tflops_per_gpu": 4.360152483412918,
    "achieved_tflops_per_gpu_theoretical": 728.3994810069811,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.06942301988601685,
    "mfu_percent": 0.4408647607090918,
    "mfu_percent_theoretical": 73.650099191808,
    "total_flos": 7.09204861186474e+17,
    "train_loss": 0.32849771289270874,
    "train_runtime": 20331.9971,
    "train_samples_per_second": 3.436,
    "train_steps_per_second": 0.215,
    "valid_targets_mean": 1305.1,
    "valid_targets_min": 490
}