train_results.json

{
    "achieved_tflops_per_gpu": 5.871788667081258,
    "achieved_tflops_per_gpu_theoretical": 528.5131118004142,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.042520757764577866,
    "mfu_percent": 0.5937096731123618,
    "mfu_percent_theoretical": 53.439141739172314,
    "total_flos": 3.824454126066991e+18,
    "train_loss": 0.1807381806187807,
    "train_runtime": 81415.8671,
    "train_samples_per_second": 2.495,
    "train_steps_per_second": 0.156,
    "valid_targets_mean": 4458.2,
    "valid_targets_min": 1268
}