fine-tune-models

fine-tune-models copied to clipboard

Could not find parameter named "kernel" in scope "/encoder/mid_block/attentions_0/query"

1

Hi, upon loading this up with any dataset and the 1.4 vae converted to jax, i get this warning on loading: The checkpoint D:\gitprojects\finetuningvae\backup\stable-diffusion-v1-5-jax\vae\ is missing required keys: {('decoder', 'mid_block',...

CiaraStrawberry

does this work with normal pytorch sd and not just stablediffusion-jax?

ezra-ch

I've implemented PatchGAN and StyleGAN Loss for decoder only finetuned, but I got artifacts reconstructions even after 50k steps. My Loss function setting is: `0.1 LPIPS + 0.1 PatchGAN (with...

mhh0318

Hi, I see your code of run_finetune_vqgan.py and Kaggle dataset, but I can't understand the dataset format of "danbooru_image_paths_ds.json". Can you give me its dataset format?

Umal-exvc