Training or fit function parameter for model updating or incremental training

[ShiyangTey]

Hi I am a new player in this field. I would like to suggest if there could be a function for a saved model to be trained on a small and new dataset to update the weights of the model, to commodate for concept/data drifting and also for more accurate model with small training data. Currently the fit function will reset the weights of the model and retrain it from scratch.

I found some examples in xgboost where the train and fit function has a parameter to continue training on a saved model [https://xgboost.readthedocs.io/en/latest/python/python_api.html#module-xgboost.sklearn].
the parameter detail is as xgb_model (file name of stored xgb model or 'Booster' instance) – Xgb model to be loaded before training (allows training continuation).

Or also like sklearn's partial fit function (I'm not too confident in how the function works).

Thank you.