Living under a rock any missed pony7. Only 4 channel VAE?
A user on Reddit is inquiring about the VAE (Variational Autoencoder) used in the "pony7" model, noting it has only 4 channels. They are comparing this to newer models that utilize 16-channel VAEs for finer details and are seeking clarification on whether the 4-channel VAE is a limitation or if there's a misunderstanding. AI