The warning can be ignored, it’s not always true that will result in a slow down. Regarding max_n_steps we should allow passing it easily as an optional kwarg. Can you open an issue on our repository for this (and feel free to open a Pull request to fix it as well): GitHub - pymc-devs/pymc: Bayesian Modeling and Probabilistic Programming in Python
Non-code wise, will the extra n_steps actually help you? It’s very inefficient already with this many steps.