train_results.json

{
    "epoch": 0.975609756097561,
    "num_input_tokens_seen": 31457280,
    "total_flos": 1.2251230144089293e+18,
    "train_loss": 0.6374592224756876,
    "train_runtime": 1520.8651,
    "train_samples_per_second": 5.175,
    "train_steps_per_second": 0.01
}