I am using XGBClassifier for a multiclass classification problem. I used GridSearchCV to tune the hyperparameters. I thought the F1 score would be the best evaluation metric for the problem. Here’s a code snippet where I am trying to optimize max_depth and min_child_weight: