train_results.json

{
    "effective_tokens_per_sec": 4156.816967960705,
    "epoch": 3.0,
    "total_flos": 3.2487544184132076e+18,
    "train_loss": 0.8073726282178277,
    "train_runtime": 15483.8831,
    "train_samples_per_second": 3.199,
    "train_steps_per_second": 0.05
}