XGBoost Early Stopping Rounds报错：fit()不识别该参数

核心原因 #

XGBoost的scikit-learn接口（如XGBRegressor）在较新版本中，已弃用将early_stopping_rounds直接作为fit()方法参数的写法，转而要求通过callbacks参数传入官方回调类实现早停逻辑。

修复代码 #

将原fit()方法中的early_stopping_rounds=50替换为callbacks参数，使用xgb.callback.EarlyStopping回调：

model.fit(
    X_train,
    y_train,
    eval_set=[(X_valid, y_valid)],
    verbose=False,
    callbacks=[xgb.callback.EarlyStopping(rounds=50, verbose=False)]
)

完整修改后的目标函数 #

def objective(trial):
    # Suggest values for hyperparameters
    params = {
        "objective": "reg:squarederror",
        "eval_metric": "rmse",
        "tree_method": "hist",  # Use hist method
        "device": "cuda",       # Specify using GPU
        "learning_rate": trial.suggest_float("learning_rate", 0.01, 0.3, log=True),
        "max_depth": trial.suggest_int("max_depth", 3, 10),
        "min_child_weight": trial.suggest_float("min_child_weight", 1, 10),
        "gamma": trial.suggest_float("gamma", 0, 1),
        "subsample": trial.suggest_float("subsample", 0.5, 1.0),
        "colsample_bytree": trial.suggest_float("colsample_bytree", 0.5, 1.0),
        "lambda": trial.suggest_float("lambda", 1e-3, 10.0, log=True),
        "alpha": trial.suggest_float("alpha", 1e-3, 10.0, log=True),
        "n_estimators": 1000  # Define n_estimators in the initialization of the model
    }

    # Initialize the model
    model = xgb.XGBRegressor(**params)

    # Train the model with early stopping callback
    model.fit(
        X_train,
        y_train,
        eval_set=[(X_valid, y_valid)],
        verbose=False,
        callbacks=[xgb.callback.EarlyStopping(rounds=50, verbose=False)]
    )

    # Predict and calculate RMSE for validation set
    preds = model.predict(X_valid)
    rmse = mean_squared_error(y_valid, preds, squared=False)

    return rmse  # Optuna minimizes this

额外排查点 #

• 版本验证：运行print(xgb.__version__)确认版本≥1.6.0（回调方式为该版本后推荐用法）。
• 环境冲突检查：避免conda与pip混合安装XGBoost，可通过pip show xgboost或conda list xgboost确认实际运行版本。
• 旧版本兼容：若需保留early_stopping_rounds参数，需回退到XGBoost 1.5.x版本，但不推荐（旧版本存在已知bug）。

内容的提问来源于stack exchange，提问作者CraigBreezey