train_results.json

{
    "effective_tokens_per_sec": 4848.179638638243,
    "epoch": 2.0,
    "total_flos": 2.55147477658973e+18,
    "train_loss": 1.3643777534166972,
    "train_runtime": 10437.131,
    "train_samples_per_second": 2.299,
    "train_steps_per_second": 0.072
}