Would would be the best approach to incrementally train gradient boosting model (system like xgboost ot lightgbm)?

As example, model can be trained based on monthly data and now we have another few month of data that we want incrementally "add" to the existing model, not to retrain model from scratch with all data.