train_results.json

{
    "achieved_tflops_per_gpu": 0.0036890217745476277,
    "achieved_tflops_per_gpu_theoretical": 572.5558711171154,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.17596597969532013,
    "mfu_percent": 0.0002607082526182069,
    "mfu_percent_theoretical": 40.463312446439254,
    "total_flos": 1518375161495552.0,
    "train_loss": 0.21164679110105047,
    "train_runtime": 25724.5561,
    "train_samples_per_second": 2.702,
    "train_steps_per_second": 0.169,
    "valid_targets_mean": 4512.6,
    "valid_targets_min": 535
}