train_results.json

{
    "epoch": 1.0,
    "total_flos": 1.0685869290894131e+17,
    "train_loss": 3.2641759352250532,
    "train_runtime": 13025.9606,
    "train_samples": 3521,
    "train_samples_per_second": 0.27,
    "train_steps_per_second": 0.068
}