train_results.json

{
    "epoch": 2.9904761904761905,
    "num_input_tokens_seen": 8039616,
    "total_flos": 3.746792787278561e+17,
    "train_loss": 0.068369100185541,
    "train_runtime": 4403.2508,
    "train_samples_per_second": 0.715,
    "train_steps_per_second": 0.044
}