train_results.json

{
    "epoch": 4.903225806451613,
    "num_input_tokens_seen": 1299120,
    "total_flos": 5.150239379108659e+16,
    "train_loss": 0.3450760219912857,
    "train_runtime": 2142.6975,
    "train_samples_per_second": 11.57,
    "train_steps_per_second": 0.089
}