final/train_results.json

{
    "epoch": 4.0,
    "total_flos": 1.6798111346589696e+16,
    "train_loss": 13.864840042896759,
    "train_runtime": 800.9126,
    "train_samples_per_second": 3.067,
    "train_steps_per_second": 0.195
}