training_state.json

{"global_step": 3000, "epoch_step": 3045, "epoch": 46, "exhausted_backends": [], "repeats": {"qinyouzhi-flux": 0}}