shirayu commited on
Commit
f060b02
1 Parent(s): 3e50d29

update scheduler setting

Browse files

As posted [here](https://www.reddit.com/r/AnimeResearch/comments/x648bp/comment/in6vkpm/?utm_source=share&utm_medium=web2x&context=3), the default scheduler does not make good pictures.

Files changed (1) hide show
  1. README.md +12 -1
README.md CHANGED
@@ -56,7 +56,18 @@ model_id = "hakurei/waifu-diffusion"
56
  device = "cuda"
57
 
58
 
59
- pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16, revision='fp16')
 
 
 
 
 
 
 
 
 
 
 
60
  pipe = pipe.to(device)
61
 
62
  prompt = "touhou hakurei_reimu 1girl solo portrait"
 
56
  device = "cuda"
57
 
58
 
59
+ pipe = StableDiffusionPipeline.from_pretrained(
60
+ model_id,
61
+ torch_dtype=torch.float16,
62
+ revision="fp16",
63
+ scheduler=DDIMScheduler(
64
+ beta_start=0.00085,
65
+ beta_end=0.012,
66
+ beta_schedule="scaled_linear",
67
+ clip_sample=False,
68
+ set_alpha_to_one=False,
69
+ ),
70
+ )
71
  pipe = pipe.to(device)
72
 
73
  prompt = "touhou hakurei_reimu 1girl solo portrait"