Tensorflow 2.0: how can I recover the model with best (lowest) validation loss after fitting?

I'm training a tensorflow 2.0 model on a training and validation set, and then want to later on recover the epoch with the lowest validation loss and evaluate it on the test set. I'm a bit confused on what is the best way to recover the model with the lowest validation loss after training. I solved this by doing the following:

1. Add the following ModelCheckpoint callback to my callbacks list

ModelCheckpoint(filepath=str(modelpath / 'best_model_{epoch:03d}-{val_loss:.6f}.h5'),
                                         monitor='val_loss',
                                         save_best_only=True)

2. Implemented the following function:

# This function gets the path to the best model and cleans any non-best model
def get_best_model_path(modelpath):

   modelpath_ = Path(modelpath)
   best_model_list = [p for p in modelpath_.iterdir()]
   best_model_list.sort()

   for modelpath in best_model_list[:-1]:
       os.remove(modelpath)

   return best_model_list[-1]

3. Then my code will be something like this

   model.fit(<fit params>)
   best_model_path = common.get_best_model(modelpath)
   best_model = tf.keras.models.load_model(best_model_path)
   evaluate(best_model, ...)

This works because ModelCheckpoint is set to save the best epoch only, so I know that the last epoch saved will be the one with (1) largest epoch number value and (2) best validation loss. Because of the way I save the file (using formater {epoch:03d}) I guarantee that it will end up at the end of the sort order when reading the file paths in get_best_model_path(modelpath)

While this works, it seems very cumbersome and I've been wondering if Tensorflow does not offer a better way to do this.