How to add trainable parameters into Tensorflow2 Keras model loss function

Question

I am trying to train an image denoiser network with Keras (Tensorflow2). For the loss function, I want to use something like (a1 * L1_loss + a2