train_results.json

{
    "effective_tokens_per_sec": 6878.842079354381,
    "epoch": 3.0,
    "total_flos": 4.1743170019314893e+18,
    "train_loss": 0.8916917941037644,
    "train_runtime": 10519.9057,
    "train_samples_per_second": 4.289,
    "train_steps_per_second": 0.134
}