Hugging Face's logo

bdsqlsz
/
stable-diffusion-xl-base-1.0_fixvae_fp16

stable-diffusion
Model card Files Community
bdsqlsz commited on
Commit
f579c63
•
1 Parent(s): 38c3c49

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +0 -5
README.md CHANGED
@@ -22,11 +22,6 @@ same vae license on sdxl-vae-fp16-fix
22
 
23
  SDXL-VAE-FP16-Fix is the [SDXL VAE](https://huggingface.co/stabilityai/sdxl-vae)*, but modified to run in fp16 precision without generating NaNs.
24
 
25
- | VAE | Decoding in `float32` / `bfloat16` precision | Decoding in `float16` precision |
26
- | --------------------- | -------------------------------------------- | ------------------------------- |
27
- | SDXL-VAE | ✅ ![](./images/orig-fp32.png) | ⚠️ ![](./images/orig-fp16.png) |
28
- | SDXL-VAE-FP16-Fix | ✅ ![](./images/fix-fp32.png) | ✅ ![](./images/fix-fp16.png) |
29
-
30
  ## Details
31
 
32
  SDXL-VAE generates NaNs in fp16 because the internal activation values are too big:
 
22
 
23
  SDXL-VAE-FP16-Fix is the [SDXL VAE](https://huggingface.co/stabilityai/sdxl-vae)*, but modified to run in fp16 precision without generating NaNs.
24
 
 
 
 
 
 
25
  ## Details
26
 
27
  SDXL-VAE generates NaNs in fp16 because the internal activation values are too big: