Retraining model but keeping the previous utterance

[stickyburn]

Hey there. Thanks for giving us such a dope framework as opensource. I have one noob question; If i have already trained my model once with say, 50% data, and now I want to add the rest 50%, how can I use the previous learnt states to build on top of that new data ? Is that even possible or will I have to retrain all over again with the new datasets ? Thanks!