train_results.json

{
    "achieved_tflops_per_gpu": 4.246655504814489,
    "achieved_tflops_per_gpu_theoretical": 711.4599581928652,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.259745717048645,
    "mfu_percent": 0.4293888275848826,
    "mfu_percent_theoretical": 71.93730618734733,
    "total_flos": 7.12082116746281e+16,
    "train_loss": 0.38846772432732746,
    "train_runtime": 2096.0086,
    "train_samples_per_second": 3.336,
    "train_steps_per_second": 0.21,
    "valid_targets_mean": 3009.9,
    "valid_targets_min": 873
}