train_results.json

{
    "effective_tokens_per_sec": 2906.7457386466012,
    "epoch": 3.0,
    "total_flos": 2.4882019125669396e+18,
    "train_loss": 1.4736498374943825,
    "train_runtime": 8380.6004,
    "train_samples_per_second": 3.746,
    "train_steps_per_second": 0.234
}