train_results.json

{
    "achieved_tflops_per_gpu": 0.003978066363362,
    "achieved_tflops_per_gpu_theoretical": 538.6768254888838,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.15846839547157288,
    "mfu_percent": 0.0002811354320397173,
    "mfu_percent_theoretical": 38.069033603454685,
    "total_flos": 1765744004825088.0,
    "train_loss": 0.17469855595608147,
    "train_runtime": 27741.8701,
    "train_samples_per_second": 2.541,
    "train_steps_per_second": 0.159,
    "valid_targets_mean": 5295.1,
    "valid_targets_min": 1258
}