train_results.json

{
    "epoch": 4.951768488745981,
    "num_input_tokens_seen": 6318560,
    "total_flos": 2.5049338265940787e+17,
    "train_loss": 0.9364173337708336,
    "train_runtime": 4331.1035,
    "train_samples_per_second": 22.95,
    "train_steps_per_second": 0.089
}