train_results.json

{
    "epoch": 2.9928263988522237,
    "num_input_tokens_seen": 2783872,
    "total_flos": 1.191738412451758e+17,
    "train_loss": 0.08308002222828938,
    "train_runtime": 963.6682,
    "train_samples_per_second": 8.673,
    "train_steps_per_second": 1.083
}