qwen72B_models/qwen_72B_d4_iter5_model/train_results.json

{
    "epoch": 2.0,
    "total_flos": 267294518280192.0,
    "train_loss": 0.3625834340879903,
    "train_runtime": 2672.7207,
    "train_samples_per_second": 2.02,
    "train_steps_per_second": 0.126
}