Hugging Face's logo

AiArtLab
/
sdxs

Text-to-Image
Diffusers
Safetensors
Model card Files
xet
Community
recoilme commited on
Commit
1fc6dba
·
verified ·
1 Parent(s): 3d8b2a1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +93 -89
README.md CHANGED
@@ -1,90 +1,94 @@
1
- ---
2
- license: apache-2.0
3
- pipeline_tag: text-to-image
4
- ---
5
-
6
- # Simple Diffusion XS
7
-
8
- *XS Size, Excess Quality*
9
-
10
- ![promo](media/promo.png)
11
-
12
- ![sdxs10](media/sdxs10.png)
13
-
14
- At AiArtLab, we strive to create a free, compact and fast model that can be trained on consumer graphics cards.
15
-
16
- - Model: 1.5b parameters
17
- - Text encoder: Qwen3 (0.6B parameters)
18
- - VAE: 8x16ch, [Simple VAE](https://huggingface.co/AiArtLab/simplevae)
19
-
20
-
21
- The model was trained for ~3 months on (2-4)x RTX 5090 GPUs using over 1+ million images of various resolutions and styles, primarily anime and illustrations.
22
-
23
- [Gradio](https://sdxs.aiartlab.org/)
24
-
25
- ### Example
26
-
27
- ```
28
- import torch
29
- from diffusers import DiffusionPipeline
30
-
31
- device = "cuda" if torch.cuda.is_available() else "cpu"
32
- dtype = torch.float16 if torch.cuda.is_available() else torch.float32
33
-
34
- pipe_id = "AiArtLab/sdxs"
35
- pipe = DiffusionPipeline.from_pretrained(
36
- pipe_id,
37
- torch_dtype=dtype,
38
- trust_remote_code=True
39
- ).to(device)
40
-
41
- prompt = "girl, smiling, red eyes, blue hair, white shirt"
42
- negative_prompt="low quality, bad quality"
43
- image = pipe(
44
- prompt=prompt,
45
- negative_prompt = negative_prompt,
46
- ).images[0]
47
-
48
- image.show(image)
49
- ```
50
-
51
- ### Model Limitations:
52
- - Limited concept coverage due to the small dataset.
53
-
54
- ## Acknowledgments
55
- - **[Stan](https://t.me/Stangle)** — Key investor. Thank you for believing in us when others called it madness.
56
- - **Captainsaturnus**
57
- - **Love. Death. Transformers.**
58
- - **TOPAPEC**
59
-
60
- ## Datasets
61
- - **[CaptionEmporium](https://huggingface.co/CaptionEmporium)**
62
-
63
- ## Donations
64
-
65
- Please contact with us if you may provide some GPU's or money on training
66
-
67
- DOGE: DEw2DR8C7BnF8GgcrfTzUjSnGkuMeJhg83
68
-
69
- BTC: 3JHv9Hb8kEW8zMAccdgCdZGfrHeMhH1rpN
70
-
71
- ## Contacts
72
-
73
- [recoilme](https://t.me/recoilme) *prefered way
74
-
75
-
76
- mail at aiartlab.org (slow response)
77
-
78
- ## More examples
79
-
80
- ![result_grid](media/result_grid.jpg)
81
-
82
- ## Citation
83
- ```bibtex
84
- @misc{sdxs,
85
- title={Simple Diffusion XS},
86
- author={recoilme with help of AiArtLab Team},
87
- url={https://huggingface.co/AiArtLab/sdxs},
88
- year={2025}
89
- }
 
 
 
 
90
  ```
 
1
+ ---
2
+ license: apache-2.0
3
+ pipeline_tag: text-to-image
4
+ ---
5
+
6
+ # Simple Diffusion XS
7
+
8
+ *XS Size, Excess Quality*
9
+
10
+ # This is a research repository for experimenting with different architectures
11
+
12
+ ![stop](https://avatars.mds.yandex.net/i?id=0b111def88aeaf6f9142cee41c573f78_l-5220579-images-thumbs&n=13)
13
+
14
+ ![promo](media/promo.png)
15
+
16
+ ![sdxs10](media/sdxs10.png)
17
+
18
+ At AiArtLab, we strive to create a free, compact and fast model that can be trained on consumer graphics cards.
19
+
20
+ - Model: 1.5b parameters
21
+ - Text encoder: Qwen3 (0.6B parameters)
22
+ - VAE: 8x16ch, [Simple VAE](https://huggingface.co/AiArtLab/simplevae)
23
+
24
+
25
+ The model was trained for ~3 months on (2-4)x RTX 5090 GPUs using over 1+ million images of various resolutions and styles, primarily anime and illustrations.
26
+
27
+ [Gradio](https://sdxs.aiartlab.org/)
28
+
29
+ ### Example
30
+
31
+ ```
32
+ import torch
33
+ from diffusers import DiffusionPipeline
34
+
35
+ device = "cuda" if torch.cuda.is_available() else "cpu"
36
+ dtype = torch.float16 if torch.cuda.is_available() else torch.float32
37
+
38
+ pipe_id = "AiArtLab/sdxs"
39
+ pipe = DiffusionPipeline.from_pretrained(
40
+ pipe_id,
41
+ torch_dtype=dtype,
42
+ trust_remote_code=True
43
+ ).to(device)
44
+
45
+ prompt = "girl, smiling, red eyes, blue hair, white shirt"
46
+ negative_prompt="low quality, bad quality"
47
+ image = pipe(
48
+ prompt=prompt,
49
+ negative_prompt = negative_prompt,
50
+ ).images[0]
51
+
52
+ image.show(image)
53
+ ```
54
+
55
+ ### Model Limitations:
56
+ - Limited concept coverage due to the small dataset.
57
+
58
+ ## Acknowledgments
59
+ - **[Stan](https://t.me/Stangle)** — Key investor. Thank you for believing in us when others called it madness.
60
+ - **Captainsaturnus**
61
+ - **Love. Death. Transformers.**
62
+ - **TOPAPEC**
63
+
64
+ ## Datasets
65
+ - **[CaptionEmporium](https://huggingface.co/CaptionEmporium)**
66
+
67
+ ## Donations
68
+
69
+ Please contact with us if you may provide some GPU's or money on training
70
+
71
+ DOGE: DEw2DR8C7BnF8GgcrfTzUjSnGkuMeJhg83
72
+
73
+ BTC: 3JHv9Hb8kEW8zMAccdgCdZGfrHeMhH1rpN
74
+
75
+ ## Contacts
76
+
77
+ [recoilme](https://t.me/recoilme) *prefered way
78
+
79
+
80
+ mail at aiartlab.org (slow response)
81
+
82
+ ## More examples
83
+
84
+ ![result_grid](media/result_grid.jpg)
85
+
86
+ ## Citation
87
+ ```bibtex
88
+ @misc{sdxs,
89
+ title={Simple Diffusion XS},
90
+ author={recoilme with help of AiArtLab Team},
91
+ url={https://huggingface.co/AiArtLab/sdxs},
92
+ year={2025}
93
+ }
94
  ```