all_results.json

{
    "epoch": 3.76,
    "train_loss": 1.082658023883899,
    "train_output_tokens": 1298884,
    "train_runtime": 22627.0931,
    "train_samples": 26000,
    "train_samples_per_second": 4.596,
    "train_steps_per_second": 0.002,
    "train_total_tokens": 9672793
}