qwen72B_models/qwen_72B_d1_iter2_model/train_results.json

{
    "epoch": 2.0,
    "total_flos": 312432690462720.0,
    "train_loss": 0.5284144356405946,
    "train_runtime": 2754.0946,
    "train_samples_per_second": 1.961,
    "train_steps_per_second": 0.123
}