train_results.json

{
    "epoch": 701.7610619469026,
    "num_input_tokens_seen": 22718312,
    "total_flos": 9.574707253860557e+16,
    "train_loss": 0.0016468607849636553,
    "train_runtime": 17323.8189,
    "train_samples_per_second": 9.236,
    "train_steps_per_second": 2.309
}