train_results.json

{
    "achieved_tflops_per_gpu": 7.528449024090184,
    "achieved_tflops_per_gpu_theoretical": 555.6163483029995,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.0703798308968544,
    "mfu_percent": 2.412964430798136,
    "mfu_percent_theoretical": 178.08216291762807,
    "total_flos": 3.5396290752681083e+18,
    "train_loss": 0.038640074608406225,
    "train_runtime": 58770.888,
    "train_samples_per_second": 2.619,
    "train_steps_per_second": 0.164,
    "valid_targets_mean": 1910.0,
    "valid_targets_min": 393
}