Thanks. This was something I tried earlier. I noticed that using the maximum number of estimators almost always lead to the highest cross validation score. I was worried there would be some overfitting as a result.