Table 12

Inconsistency due to changes in the Variational Autoencoder (VAE) for the Kandinsky diffusion model and with EfficientNet v2small model.

Testing/TrainingKandinsky 2.2Autoencoder with KL Loss (AKL) 1.5Autoencoder with KL Loss (AKL) 2Consistency DecoderVAE from Pixart-αVAE from SD 2.1VAE from SD TurboοVAE from SD XL TurboSD1.4 EMA fine-tuned VAESD1.4 MSE fine-tuned VAESDXL fine-tuned VAETiny AutoEncoder for SD
Kandinsky 2.21.26%1.41%1.40%1.33%1.39%1.42%1.37%1.37%1.40%1.33%1.43%1.44%

or Create an Account

Close Modal
Close Modal