---
license: apache-2.0
---

## EQ-VAE: Equivariance Regularized Latent Space for Improved Generative Image Modeling
Arxiv: https://arxiv.org/abs/2502.09509 <br>


**EQ-VAE** regularizes the latent space of pretrained autoencoders by enforcing equivariance under scaling and rotation transformations. 

---
#### Model Description 
This model is [SiT-XL/2-REPA](https://sihyun.me/REPA/) trained with latents extracted with [EQ-VAE](https://huggingface.co/zelaki/eq-vae). The model is trained for 1M iterations on Imagenet 256x256.