Deep learning with Flux.jl and gumbel-softmax trick

I am new to Julia, but I’m evaluating Flux.jl as an option to implement a Variational Autoencoder (VAE). But, unfortunately, I’m not finding an example using the gumbel-softmax trick. I found on Pytorch, but I wonder how could I make it work with Flux.jl. Could someone give me a direction on this?

1 Like

I’ve never heard of a Julia library called Flow.jl. Can you link the one you’re evaluating and the examples you’ve looked through which are relevant but don’t include gumbel softmax?

1 Like

I am running this example with Fashion MNIST dataset: Convolutional VAE in Flux | Aleco Kastanos

But in this case is a classical VAE with the Normal distribution. I don’t know how could I adapt it to use the gumbel-softmax trick.