train_results.json

{
    "epoch": 20.0,
    "num_input_tokens_seen": 221910640,
    "total_flos": 8.056851732185088e+16,
    "train_loss": 0.641815161904031,
    "train_runtime": 7964.4512,
    "train_samples": 153338,
    "train_samples_per_second": 385.056,
    "train_steps_per_second": 48.134,
    "train_tokens_per_second": 27853.103
}