how to improve the Accuracy by fine-tuning the parameter of gcForest？

Hi, I am sorry about the previous question. the gcForest was working well under my environment. However, I am confused about the result after using gcForest to handle the multi-classification problem. Here is the related code of my issue.

 def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument("--model", dest="model", type=str, default=None, help="gcfoest Net Model File")
    args = parser.parse_args()
    return args


def get_toy_config():
config = {}
ca_config = {}
ca_config["random_state"] = 0
ca_config["max_layers"] = 100
ca_config["early_stopping_rounds"] = 3
ca_config["n_classes"] = 3
ca_config["estimators"] = []
ca_config["estimators"].append(
        {"n_folds": 5, "type": "XGBClassifier", "n_estimators": 10, "max_depth": 5,
         "objective": "multi:softprob", "silent": True, "nthread": -1, "learning_rate": 0.1} )
ca_config["estimators"].append({"n_folds": 5, "type": "RandomForestClassifier", "n_estimators": 10, "max_depth": None, "n_jobs": -1})
ca_config["estimators"].append({"n_folds": 5, "type": "ExtraTreesClassifier", "n_estimators": 10, "max_depth": None, "n_jobs": -1})
ca_config["estimators"].append({"n_folds": 5, "type": "LogisticRegression"})
config["cascade"] = ca_config
return config

In my case, I just change the number of classes to three. BTW, the input vector was the matrix（1280*320)， and the labeled data was the matrix（1280,) . It turns out the accuracy of leave-one-group-out was just like this. And, I used the MLP for my data also. the result is much better than gcForest. Do you have any clue for this problem？maybe the hyper-parameter of gcForest？ Thanks for your patience. Best regards Irving