Re-train a saved tensorflow model for additional labels

Question

In the transfer learning example

https://www.tensorflow.org/tutorials/images/transfer_learning_with_hub

Is it possible to train on a certain number of flower images and save it for predictions and later load the saved model and train on a new set of flowers and be able to predict both the type of flowers

ie.

Train on flowers labels and data for [daisy, dandelion] save the model say model_d prediction works for daisy, dandelion

Load the saved model model_d Train on flowers labels and data for [roses, tulips] Would prediction work for [daisy, dandelion, roses , tulips]

ie do it in small batches but the model gains more labels over many training iterations

Any example on this would be helpful