train_results.json

{
    "epoch": 4.887459807073955,
    "num_input_tokens_seen": 1299392,
    "total_flos": 5.151317702790349e+16,
    "train_loss": 0.3433768034317166,
    "train_runtime": 2162.0959,
    "train_samples_per_second": 11.489,
    "train_steps_per_second": 0.088
}