How to reset the learning rate in Tensorflow 2

Question

I used tf.keras.callbacks.ReduceLROnPlateau to reduce the learning rate during the training and then saved the weights of my model. After I load the weights again, it seems Tensorflow remembers the last learning rate used. I tried to reset the learning rate in the optimizer's definition bytf.keras.optimizers.RMSprop(learning_rate = 0.001), however this does not work and the Tensorflow still uses the last used learning rate. Then I tried to set the minimum learning rate in tf.keras.callbacks.ReduceLROnPlateauto what I want but this does not work either. How can I change the learning rate of a model that I saved its weights?