train_results.json

{
    "epoch": 1.9884169884169884,
    "num_input_tokens_seen": 12333744,
    "total_flos": 2.0534009572779622e+17,
    "train_loss": 0.535006606301596,
    "train_runtime": 453.6685,
    "train_samples_per_second": 9.13,
    "train_steps_per_second": 0.569
}