Can't use combination of gradiants for multiple losses functions of a multi-output keras model

Question

I am doing a time-series forecasting in Keras with a CNN and the EHR dataset. The goal is to predict both what molecule to give to the patient and the time until the next patien