train_results.json

{
    "achieved_tflops_per_gpu": 3.8365622148613308,
    "achieved_tflops_per_gpu_theoretical": 121.96456426163404,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.42960670590400696,
    "mfu_percent": 0.271135138859458,
    "mfu_percent_theoretical": 8.619403834744455,
    "total_flos": 1.3335511329281147e+18,
    "train_loss": 0.5088972483470942,
    "train_runtime": 21724.3827,
    "train_samples_per_second": 2.299,
    "train_steps_per_second": 0.036,
    "valid_targets_mean": 3849.6,
    "valid_targets_min": 730
}