adding partial charges about chemprop

[swansonk14]

Thank you for the suggestion. We have deprecated this repo in favor of https://github.com/swansonk14/chemprop, but we plan to add partial charge as a feature in the near future.

[yangkevin2]

Hi,

I tried the suggested change on bace, freesolv, and delaney random splits over 10 folds and it made no difference in performance. In addition, the features caused the model to get nans on several datasets I tested (tox21, lipo, sider), since I guess the partial charges computation might not work sometimes? Anyway since it makes no difference in performance I'm going to close the issue and not push this change for now. Thank you for the suggestion though; it was definitely worth a try.