train_results.json

{
    "epoch": 2.989247311827957,
    "num_input_tokens_seen": 1855776,
    "total_flos": 7.944329136203366e+16,
    "train_loss": 0.08561765917459097,
    "train_runtime": 638.811,
    "train_samples_per_second": 8.721,
    "train_steps_per_second": 1.09
}