Dlux804 / McQuade-Chem-ML

Development of easy to use and reproducible ML scripts for chemistry.
5 stars 1 forks source link

Save Sklearn Machine Learning Results #34

Closed qle2 closed 3 years ago

qle2 commented 4 years ago

Is your feature request related to a problem? Please describe. The current sklearn machine learning workflow doesn't save machine learning models besides outputs such as scores and graphs.

Describe the solution you'd like Add a way to save and load machine learning models using Pickle or joblib.

Additional context https://machinelearningmastery.com/save-load-machine-learning-models-python-scikit-learn/

qle2 commented 4 years ago

Done in Pull Request #47

Dlux804 commented 4 years ago

I argue that this is not yet closed. You have implemented saving the models via pkl files output by the Bayesian optimizer, but I do not recall a method for loading models in your last PR. Additional question -- what is being saved in the pkl file from Bayes? At that point, I am not sure we have our final model yet. We receive from the Bayes optimizer the optimal parameters, but I believe we need to retrain the model with those parameters (they were not necessarily the last ones used in the Bayes algorithm.)

Additionally, if we do not tune the hyper parameters, the model will never be stored.