Are there speed and memory benchmarks (e.g. compared to scikit-learn, obliquetree etc.) for overfitting a dataset completely (no limit on tree depth)? This is usually a fair comparison for speed and memory without having to care about the accuracy part.