I have been having this issue also. Conda may well be better, but how do you use conda in databricks? From what I can see the only ways are to either use runtime 7.3 or below or use completely new type of container. Those are not feasible.