train_results.json

{
    "effective_tokens_per_sec": 4038.2883152185364,
    "epoch": 4.0,
    "total_flos": 3.2734142949973033e+18,
    "train_loss": 1.3248716153500641,
    "train_runtime": 16065.8623,
    "train_samples_per_second": 2.605,
    "train_steps_per_second": 0.081
}