train_results.json

{
    "epoch": 2.976,
    "num_input_tokens_seen": 3930992,
    "total_flos": 3.112338516554547e+16,
    "train_loss": 0.7158093615885703,
    "train_runtime": 1518.3826,
    "train_samples_per_second": 3.952,
    "train_steps_per_second": 0.122
}