benchmark/train_results.json

{
    "epoch": 11.389830508474576,
    "num_input_tokens_seen": 1664720,
    "total_flos": 7.517105945247744e+16,
    "train_loss": 1.1265570782124996,
    "train_runtime": 1670.3142,
    "train_samples_per_second": 0.841,
    "train_steps_per_second": 0.101
}