Hugging Face's logo

kandinskylab
/
Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers

Diffusers
Safetensors
Kandinsky5T2VPipeline
Model card Files
xet
 Community
7

Update README.md

#4
by leffff - opened
base: refs/heads/main
←
from: refs/pr/4
Discussion Files changed
+5
-5
Files changed (1) hide show
  1. README.md +5 -5
README.md CHANGED
@@ -43,7 +43,7 @@ from diffusers.utils import export_to_video
43
  # Load the pipeline
44
  pipe = Kandinsky5T2VPipeline.from_pretrained(
45
  "ai-forever/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers",
46
- torch_dtype=torch.float16
47
  )
48
  pipe = pipe.to("cuda")
49
 
@@ -73,25 +73,25 @@ from diffusers import Kandinsky5T2VPipeline
73
  # SFT 5s model (highest quality)
74
  pipe_sft = Kandinsky5T2VPipeline.from_pretrained(
75
  "ai-forever/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers",
76
- torch_dtype=torch.float16
77
  )
78
 
79
  # Distilled 16-step model (fastest)
80
  pipe_distill = Kandinsky5T2VPipeline.from_pretrained(
81
  "ai-forever/Kandinsky-5.0-T2V-Lite-distilled16steps-5s-Diffusers",
82
- torch_dtype=torch.float16
83
  )
84
 
85
  # No-CFG model (balanced speed/quality)
86
  pipe_nocfg = Kandinsky5T2VPipeline.from_pretrained(
87
  "ai-forever/Kandinsky-5.0-T2V-Lite-nocfg-5s-Diffusers",
88
- torch_dtype=torch.float16
89
  )
90
 
91
  # Pretrain model (most diverse)
92
  pipe_pretrain = Kandinsky5T2VPipeline.from_pretrained(
93
  "ai-forever/Kandinsky-5.0-T2V-Lite-pretrain-5s-Diffusers",
94
- torch_dtype=torch.float16
95
  )
96
  ```
97
 
 
43
  # Load the pipeline
44
  pipe = Kandinsky5T2VPipeline.from_pretrained(
45
  "ai-forever/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers",
46
+ torch_dtype=torch.bfloat16
47
  )
48
  pipe = pipe.to("cuda")
49
 
 
73
  # SFT 5s model (highest quality)
74
  pipe_sft = Kandinsky5T2VPipeline.from_pretrained(
75
  "ai-forever/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers",
76
+ torch_dtype=torch.bfloat16
77
  )
78
 
79
  # Distilled 16-step model (fastest)
80
  pipe_distill = Kandinsky5T2VPipeline.from_pretrained(
81
  "ai-forever/Kandinsky-5.0-T2V-Lite-distilled16steps-5s-Diffusers",
82
+ torch_dtype=torch.bfloat16
83
  )
84
 
85
  # No-CFG model (balanced speed/quality)
86
  pipe_nocfg = Kandinsky5T2VPipeline.from_pretrained(
87
  "ai-forever/Kandinsky-5.0-T2V-Lite-nocfg-5s-Diffusers",
88
+ torch_dtype=torch.bfloat16
89
  )
90
 
91
  # Pretrain model (most diverse)
92
  pipe_pretrain = Kandinsky5T2VPipeline.from_pretrained(
93
  "ai-forever/Kandinsky-5.0-T2V-Lite-pretrain-5s-Diffusers",
94
+ torch_dtype=torch.bfloat16
95
  )
96
  ```
97