llama-factory/saves/Qwen2-72B-Instruct/train_results.json

{
    "epoch": 5.997322623828648,
    "total_flos": 1.7687758754493235e+18,
    "train_loss": 0.5969825589208908,
    "train_runtime": 20472.2071,
    "train_samples_per_second": 1.314,
    "train_steps_per_second": 0.082
}