train_results.json

{
    "epoch": 0.9984,
    "num_input_tokens_seen": 5111808,
    "total_flos": 2.1873318928633037e+17,
    "train_loss": 0.5760276004289969,
    "train_runtime": 1768.2955,
    "train_samples_per_second": 5.655,
    "train_steps_per_second": 0.088
}