Separate optimizer for discriminator and the rest of the model

1

I have a model with the following components.

  1. embedding layer
  2. encoder
  3. generator
  4. discriminator
  5. feed-forward neural network

I want to define two optimizers. One for the discriminator only and one for the rest. I am doing the following.

optimizers = []
model_params = chain(model.embedding.parameters(), model.encoder.parameters(), 
                            model.generator.parameters(), model.ffnn.parameters())
optimizers.append(optim.Adam(model_params, args.lr))
optimizers.append(optim.Adam(model.discriminator.parameters(), args.lr))

Is there any better way to do the same? For example, can I take the difference between model.parameters() and model.discriminator.parameters()? If yes, how can I do that?