all_results.json

{
    "epoch": 2.96,
    "num_input_tokens_seen": 358624,
    "total_flos": 1.5352214341287936e+16,
    "train_loss": 0.24338918279957128,
    "train_runtime": 8119.2255,
    "train_samples_per_second": 0.111,
    "train_steps_per_second": 0.014
}