I am trying to follow the hyperparameter search guide to limit the number of iterations, but not seeing it work. Rather it is going through all the parameters. For example, my understanding from the docs is to use random search in order to limit the iterations set in search_args as shown:

SELECT * FROM pgml.train(
    'Handwritten Digit Image Classifier', 
    algorithm => 'xgboost', 
    search => 'random', 
    search_args => '{ "n_iter": 2 }',
    search_params => '{
        "max_depth": [1, 2, 3, 4, 5, 6], 
        "n_estimators": [20, 40, 80, 160]
    }'
);

However, I am seeing it iterate through all the search params rather than limited to just 2 iterations. What am I doing wrong or misunderstanding?