train_results.json

{
    "epoch": 20.0,
    "num_input_tokens_seen": 24887720,
    "total_flos": 1.1207225397864038e+18,
    "train_loss": 0.4843763030143904,
    "train_runtime": 6609.4499,
    "train_samples_per_second": 23.085,
    "train_steps_per_second": 5.774
}