BlinkDL commited on
Commit
bc54ee8
·
1 Parent(s): 6f9f76f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -4
README.md CHANGED
@@ -4,10 +4,8 @@ license: apache-2.0
4
 
5
  This model can encode 224x224 RGB image into 28x28x13bit (1274 bytes) latent. The compression rate is 28x28x13/(224x224x24)=1/118, or 0.203 bpp (same as VQGAN_f8_8192).
6
 
7
- 12M params for Encoder + Decoder. Trained on LAION-Aesthetics V2 5+ for 60M images.
8
 
9
- Guided by https://huggingface.co/laion/CLIP-ViT-B-32-laion2B-s34B-b79K (it's great. better than OpenAI CLIP B/32) and https://github.com/dingkeyan93/DISTS.
10
-
11
- No GAN loss. So probably the image is slightly blurred in some cases?
12
 
13
  (still training. final checkpt will be better)
 
4
 
5
  This model can encode 224x224 RGB image into 28x28x13bit (1274 bytes) latent. The compression rate is 28x28x13/(224x224x24)=1/118, or 0.203 bpp (same as VQGAN_f8_8192).
6
 
7
+ 12M params for Encoder + Decoder. Trained on LAION-Aesthetics V2 5+ for 130M images.
8
 
9
+ Guided by https://huggingface.co/laion/CLIP-ViT-B-32-laion2B-s34B-b79K (it's great. better than OpenAI CLIP B/32) and https://github.com/dingkeyan93/DISTS. No GAN loss.
 
 
10
 
11
  (still training. final checkpt will be better)