train_results.json

{
    "epoch": 128.0,
    "total_flos": 1.5995217836155625e+19,
    "train_loss": 1.3990157055003303,
    "train_runtime": 26333.2616,
    "train_samples": 448,
    "train_samples_per_second": 2.552,
    "train_steps_per_second": 0.08
}