train_results.json

{
    "epoch": 0.9966329966329966,
    "num_input_tokens_seen": 3054720,
    "total_flos": 4.410158569555968e+16,
    "train_loss": 0.5683108514618456,
    "train_runtime": 2363.4699,
    "train_samples_per_second": 1.257,
    "train_steps_per_second": 0.078
}