Search Results
7/22/2025, 5:54:15 PM
>>105988677
EQ-VAE is the savior. He's fighting the VAE and it seems to very much underrepresent nsfw, cartoons, anime, etc, and the model has to deal with lots of noise. Basically it's like having a really bad, noisy dataset except at the latents level. In this example you have to see that the model has to reconcile the noise within the VAE itself.
EQ-VAE is the savior. He's fighting the VAE and it seems to very much underrepresent nsfw, cartoons, anime, etc, and the model has to deal with lots of noise. Basically it's like having a really bad, noisy dataset except at the latents level. In this example you have to see that the model has to reconcile the noise within the VAE itself.
Page 1