Remember about the original post.

_posts/2023-04-21-denoising-diffusion-impl/denoising-diffusion-impl.Rmd

@@ -36,6 +36,7 @@ reference papers.
 We are going to use `luz` to train the denoising diffusion model. Using luz allows
 us to avoid much of the torch boilerplate code related to supporting training in different
 devices like the cpu or the GPU, logging metrics, disabling autograd during evaluation, etc.
+Our implementation is heavily inspired by the [Denoising Diffusion Implicit Models](https://keras.io/examples/generative/ddim/) example in the Keras gallery.
 
 However, since we are not building a standard supervised model, we will need to 
 implement a custom `step` method.