Hi! Thanks for the wonderful work and releasing the code!
I am a bit confused by the eq-vae and eq-vae-ema that you released - what specfication do these model corrsponding to the models listed in paper Appendix Table 13?
i.e.,
Is eq-vae initialized from SD-VAE and then tuned with EQ on OpenImages?
Is eq-vae-ema initialized from SD-VAE-FT-EMA and then tuned with EQ on ImageNet?
And I also noticed that the size of weights (.ckpt vs .safetensors) that you released in https://huggingface.co/zelaki/eq-vae-ldm/tree/main and https://huggingface.co/zelaki/eq-vae/tree/main are different. But they are for the same model?
WIll you consider to release other EQ-VAE model checkpoints?
Thank you very much for your time and help!
Hi! Thanks for the wonderful work and releasing the code!
I am a bit confused by the eq-vae and eq-vae-ema that you released - what specfication do these model corrsponding to the models listed in paper Appendix Table 13?
i.e.,
Is eq-vae initialized from SD-VAE and then tuned with EQ on OpenImages?
Is eq-vae-ema initialized from SD-VAE-FT-EMA and then tuned with EQ on ImageNet?
And I also noticed that the size of weights (.ckpt vs .safetensors) that you released in https://huggingface.co/zelaki/eq-vae-ldm/tree/main and https://huggingface.co/zelaki/eq-vae/tree/main are different. But they are for the same model?
WIll you consider to release other EQ-VAE model checkpoints?
Thank you very much for your time and help!