train_results.json

{
    "epoch": 83.16008316008316,
    "num_input_tokens_seen": 30508240,
    "total_flos": 1.3737722584517837e+18,
    "train_loss": 0.2765025964975357,
    "train_runtime": 24417.94,
    "train_samples_per_second": 26.21,
    "train_steps_per_second": 1.638
}