Finetune repo: lavinal712/AutoencoderKL
Dataset: CortexLM/midjourney-v6
Fine-tuning modules: decoder and post_quant_conv
Input:
Reconstruction:
| metrics on ImageNet | rFID | PSNR | SSIM | LPIPS |
|---|---|---|---|---|
| sd-vae-ft-mse | 0.692 | 26.910 | 0.772 | 0.130 |
| finetuned (ours) | 1.638 | 27.046 | 0.785 | 0.126 |