Why is Latent W dimensionality (1,16,512) and not (1,18,512)

[SleyAI]

Hello, I'm wondering why the dimensionality of the latent W is (1,16,512) instead of (1,18,512) which is proposed by the original StyleGAN2. I'm asking because I'm trying to edit real images which are projected into the latent space by Encoder4Editing. However these latents have a dimensionality of (1,18,512) but the random images created in your jupyter notebook have a dimensionality of (1,16,512). Is there a way to convert the latent?

[shartoo]

dimensionality of the latent W is (1,16,512) when the image resolution of generator is 512x512,and 1,18,512 for 1024x1024