all_results.json

{
    "epoch": 6.25,
    "total_flos": 4.798362622332561e+18,
    "train_loss": 0.7259744338989258,
    "train_runtime": 4809.6219,
    "train_samples": 2554,
    "train_samples_per_second": 3.327,
    "train_steps_per_second": 0.208
}