train_results.json

{
    "achieved_tflops_per_gpu": 0.0019812134097001344,
    "achieved_tflops_per_gpu_theoretical": 1033.3108916610663,
    "epoch": 6.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.12042035162448883,
    "mfu_percent": 0.00014001508195760667,
    "mfu_percent_theoretical": 73.02550471102941,
    "total_flos": 254043172110336.0,
    "train_loss": 0.18868140803068803,
    "train_runtime": 8014.1282,
    "train_samples_per_second": 4.869,
    "train_steps_per_second": 0.305,
    "valid_targets_mean": 1261.1,
    "valid_targets_min": 857
}