qwen72B_models/qwen_72B_d2_iter3_model/train_results.json

{
    "epoch": 2.0,
    "total_flos": 293955502604288.0,
    "train_loss": 0.4355149036328468,
    "train_runtime": 2715.6817,
    "train_samples_per_second": 1.988,
    "train_steps_per_second": 0.124
}