train_results.json

{
    "achieved_tflops_per_gpu": 0.0033304228836005876,
    "achieved_tflops_per_gpu_theoretical": 1351.773920566444,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.1837315857410431,
    "mfu_percent": 0.00023536557481276238,
    "mfu_percent_theoretical": 95.53172583508439,
    "total_flos": 443336046870528.0,
    "train_loss": 0.22974560444365072,
    "train_runtime": 8319.8152,
    "train_samples_per_second": 6.378,
    "train_steps_per_second": 0.399,
    "valid_targets_mean": 1887.4,
    "valid_targets_min": 635
}