train_results.json

{
    "epoch": 3.99728014505893,
    "num_input_tokens_seen": 11278520,
    "total_flos": 7.651349314204147e+17,
    "train_loss": 1.351207665421746,
    "train_runtime": 8552.2085,
    "train_samples_per_second": 2.578,
    "train_steps_per_second": 0.103
}