Eugeoter commited on
Commit
54c94a0
1 Parent(s): 5c9834d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +154 -85
README.md CHANGED
@@ -3,62 +3,119 @@ license: other
3
  license_name: fair-ai-public-license-1.0-sd
4
  license_link: https://freedevproject.org/faipl-1.0-sd/
5
  language:
6
- - en
7
  base_model:
8
- - Laxhar/noobai-XL_v1.0
9
  pipeline_tag: text-to-image
10
  tags:
11
- - Diffusers
12
- - Safetensors
 
 
 
 
13
  ---
14
 
 
 
15
  # Model Introduction
16
 
17
- This image generation model, based on Laxhar/noobai-XL_v1.0, leverages full Danbooru and e621 datasets with native tags and natural language captioning.
18
 
19
- Implemented as a v-prediction model (distinct from eps-prediction), it requires specific parameter configurations - detailed in following sections.
20
 
21
  Special thanks to my teammate euge for the coding work, and we're grateful for the technical support from many helpful community members.
22
 
23
-
24
  # ⚠️ IMPORTANT NOTICE ⚠️
25
- ## **THIS MODEL WORKS DIFFERENT FROM EPS MODELS!**
26
- ## **PLEASE READ THE GUIDE CAREFULLY!**
27
 
 
28
 
 
29
 
30
  ## Model Details
 
31
  - **Developed by**: [Laxhar Lab](https://huggingface.co/Laxhar)
32
-
33
  - **Model Type**: Diffusion-based text-to-image generative model
34
-
35
  - **Fine-tuned from**: Laxhar/noobai-XL_v1.0
36
-
37
  - **Sponsored by from**: [Lanyun Cloud](https://cloud.lanyun.net)
 
38
  ---
 
39
  # How to Use the Model.
40
 
41
- ## stable-diffusion-webui
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
42
 
43
- 1. Clone the repository
44
  ```bash
45
  git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui
46
  ```
47
 
48
- 2. Switch to dev branch
 
49
  ```bash
50
  git switch dev
51
  ```
52
 
53
- 3. Pull latest updates
 
54
  ```bash
55
  git pull
56
  ```
57
 
58
- 4. Use normally)
59
- - Follow standard procedures to launch and use the model
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
60
 
61
- **Note**: Make sure Git is installed and environment is properly configured
62
 
63
  ---
64
 
@@ -66,71 +123,89 @@ git pull
66
 
67
  ## Parameters
68
 
69
- (For vpred model, recommend using low cfg and more Steps)
70
  - CFG: 4 ~ 5
71
- - Steps: 28 ~ 35
72
- - Sampling Method:Euler a
73
- - Resolution:aim for around 1024\*1024
74
 
75
  ## Prompts
76
- - Prompt:
 
 
77
  ```
78
  masterpiece, best quality, newest, absurdres, highres, safe,
79
  ```
 
80
  - Negative Prompt:
 
81
  ```
82
- nsfw,worst quality,old,early,low quality,quality,lowres,signature,username,bad id,bad twitter id,english commentary,logo,bad hands,mutated hands,mammal,anthro,furry,ambiguous_form,feral,semi-anthro
83
  ```
84
 
85
  # Usage Guidelines
86
 
87
  ## Caption
88
- ```
89
- <1girl/1boy/1other/...>, <character>, <series>, <artists>, <special tags>, <general tags>, <other tags>
90
- ```
 
91
 
92
  ## Quality Tags
93
- For quality tags, we evaluated image popularity through the following process:
94
- - Data normalization based on various sources and ratings.
95
- - Application of time-based decay coefficients according to date recency.
96
- - Ranking of images within the entire dataset based on this processing.
97
-
 
 
98
  Our ultimate goal is to ensure that quality tags effectively track user preferences in recent years.
99
 
100
- | Percentile Range | Quality Tags |
101
- |:-----------------|:------------------|
102
- | > 95th | masterpiece |
103
- | > 85th, <= 95th | best quality |
104
- | > 60th, <= 85th | good quality |
105
- | > 30th, <= 60th | normal quality |
106
- | <= 30th | worst quality |
107
-
108
- ## Date tags
109
- | Year Range | Period |
110
- |:-----------------|:------------------|
111
- | 2005-2010 | old |
112
- | 2011-2014 | early |
113
- | 2014-2017 | mid |
114
- | 2018-2020 | recent |
115
- | 2021-2024 | newest |
116
-
117
- ## Datasets
118
- - Latest Danbooru images up to the training date(approximately before 2024-10-23)
 
 
 
 
 
 
 
 
 
 
 
 
119
  - E621 images [e621-2024-webp-4Mpixel](https://huggingface.co/datasets/NebulaeWis/e621-2024-webp-4Mpixel) dataset on Hugging Face
120
 
121
  **Communication**
122
 
123
- * **QQ Groups:**
124
- * 875042008
125
- * 914818692
126
- * 635772191
127
 
128
- * **Discord:** [Laxhar Dream Lab SDXL NOOB](https://discord.com/invite/DKnFjKEEvH)
 
 
129
 
 
130
 
131
  # Model License
132
 
133
- This model's license inherits from https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0 fair-ai-public-license-1.0-sd and adds the following terms. Any use of this model and its variants is bound by this license.
134
 
135
  ## I. Usage Restrictions
136
 
@@ -158,33 +233,27 @@ Generated models may produce unexpected or harmful outputs. Users must assume al
158
 
159
  ## Participants
160
 
161
- * **L_A_X:** [Civitai](https://civitai.com/user/L_A_X) | [Liblib.art](https://www.liblib.art/userpage/9e1b16538b9657f2a737e9c2c6ebfa69)
162
- * **li_li:** [Civitai](https://civitai.com/user/li_li)
163
- * **nebulae:** [Civitai](https://civitai.com/user/kitarz)
164
- * **Chenkin:** [Civitai](https://civitai.com/user/Chenkin)
165
- * **Euge:** [Civitai](https://civitai.com/user/Euge_)
166
-
167
 
168
  ## Contributors
169
 
170
- * **Narugo1992:** Thanks to narugo1992 and the deepghs team for open-sourcing various training sets, image processing tools, and models.
171
- * [GitHub](https://github.com/narugo1992)
172
- * [Hugging Face](https://huggingface.co/deepghs)
173
-
174
- * **Naifu:** Training scripts
175
- * [GitHub](https://github.com/Mikubill/naifu)
176
-
177
- * **Onommai:** Thanks to onommai for open-sourcing a powerful base model.
178
- * [Onommai](https://onomaai.com/)
179
-
180
- * **aria1th261:** [Civitai](https://civitai.com/user/aria1th261)
181
- * **neggles:** [GitHub](https://github.com/neggles/neurosis)
182
- * **parsee-mizuhashi:** [Hugging Face](https://huggingface.co/parsee-mizuhashi)
183
- * **bluvoll:** [Civitai](https://civitai.com/user/bluvoll)
184
- * **sdtana:** [Hugging Face](https://huggingface.co/sdtana)
185
- * **chewing:** [Hugging Face](https://huggingface.co/chewing)
186
- * **irldoggo:** [GitHub](https://github.com/irldoggo)
187
- * **reoe:** [Hugging Face](https://huggingface.co/reoe)
188
- * **kblueleaf:** [Civitai](https://civitai.com/user/kblueleaf)
189
- * **Yidhar:** [GitHub](https://github.com/Yidhar)
190
- * **Others:** ageless, 白玲可, Creeper, KaerMorh, 吟游诗人, SeASnAkE, [zwh20081](https://civitai.com/user/zwh20081), Wenaka⁧~喵, 稀里哗啦, 幸运二副, 昨日の約, 445, [EBIX](https://civitai.com/user/EBIX), [Sopp](https://huggingface.co/goyishsoyish), [Y_X](https://civitai.com/user/Y_X), adsfssdf, [Minthybasis](https://civitai.com/user/Minthybasis), [Rakosz](https://civitai.com/user/Rakosz)
 
3
  license_name: fair-ai-public-license-1.0-sd
4
  license_link: https://freedevproject.org/faipl-1.0-sd/
5
  language:
6
+ - en
7
  base_model:
8
+ - Laxhar/noobai-XL_v1.0
9
  pipeline_tag: text-to-image
10
  tags:
11
+ - safetensors
12
+ - diffusers
13
+ - stable-diffusion
14
+ - stable-diffusion-xl
15
+ - art
16
+ library_name: diffusers
17
  ---
18
 
19
+ <h1 align="center"><strong style="font-size: 48px;">NoobAI XL V-Pred 0.5</strong></h1>
20
+
21
  # Model Introduction
22
 
23
+ This image generation model, based on Laxhar/noobai-XL_v1.0, leverages full Danbooru and e621 datasets with native tags and natural language captioning.
24
 
25
+ Implemented as a v-prediction model (distinct from eps-prediction), it requires specific parameter configurations - detailed in following sections.
26
 
27
  Special thanks to my teammate euge for the coding work, and we're grateful for the technical support from many helpful community members.
28
 
 
29
  # ⚠️ IMPORTANT NOTICE ⚠️
 
 
30
 
31
+ ## **THIS MODEL WORKS DIFFERENT FROM EPS MODELS!**
32
 
33
+ ## **PLEASE READ THE GUIDE CAREFULLY!**
34
 
35
  ## Model Details
36
+
37
  - **Developed by**: [Laxhar Lab](https://huggingface.co/Laxhar)
 
38
  - **Model Type**: Diffusion-based text-to-image generative model
 
39
  - **Fine-tuned from**: Laxhar/noobai-XL_v1.0
 
40
  - **Sponsored by from**: [Lanyun Cloud](https://cloud.lanyun.net)
41
+
42
  ---
43
+
44
  # How to Use the Model.
45
 
46
+ ## Method I: [reForge](https://github.com/Panchovix/stable-diffusion-webui-reForge/tree/dev_upstream)
47
+
48
+ 1. Install reForge by following the instructions in the repository;
49
+ 2. Switch to `dev_upstream_experimental` branch by running `git checkout dev_upstream_experimental`;
50
+ 3. Launch reForge WebUI;
51
+ 4. Find "_Advanced Model Sampling for Forge_" accordion at the bottom of the "_txt2img_" tab;
52
+ 5. Enable "_Enable Advanced Model Sampling_";
53
+ 6. Select "_v_prediction_" in the "_Discrete Sampling Type_" checkbox group.
54
+ 7. Generate images!
55
+
56
+ ## Method II: [ComfyUI](https://github.com/comfyanonymous/ComfyUI)
57
+
58
+ TODO
59
+
60
+ ## Method III: [WebUI](https://github.com/AUTOMATIC1111/stable-diffusion-webui)
61
+
62
+ Note that dev branch is not stable and **may contain bugs**.
63
+
64
+ 1. (If you haven't installed WebUI) Clone the repository:
65
 
 
66
  ```bash
67
  git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui
68
  ```
69
 
70
+ 2. Switch to `dev` branch:
71
+
72
  ```bash
73
  git switch dev
74
  ```
75
 
76
+ 3. Pull latest updates:
77
+
78
  ```bash
79
  git pull
80
  ```
81
 
82
+ ## Method IV: [Diffusers](https://huggingface.co/docs/diffusers/en/index)
83
+
84
+ ```python
85
+ import torch
86
+ from diffusers import StableDiffusionXLPipeline
87
+ from diffusers import EulerDiscreteScheduler
88
+
89
+ ckpt_path = "/path/to/model.safetensors"
90
+ pipe = StableDiffusionXLPipeline.from_single_file(
91
+ ckpt_path,
92
+ use_safetensors=True,
93
+ torch_dtype=torch.float16,
94
+ )
95
+ scheduler_args = {"prediction_type": "v_prediction", "rescale_betas_zero_snr": True}
96
+ pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, **scheduler_args)
97
+ pipe.enable_xformers_memory_efficient_attention()
98
+ pipe = pipe.to("cuda")
99
+
100
+ prompt = """masterpiece, best quality,artist:john_kafka,artist:nixeu,artist:quasarcake, chromatic aberration, film grain, horror \(theme\), limited palette, x-shaped pupils, high contrast, color contrast, cold colors, arlecchino \(genshin impact\), black theme, gritty, graphite \(medium\)"""
101
+ negative_prompt = "nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro"
102
+
103
+ image = pipe(
104
+ prompt=prompt,
105
+ negative_prompt=negative_prompt,
106
+ width=832,
107
+ height=1216,
108
+ num_inference_steps=28,
109
+ guidance_scale=5,
110
+ generator=torch.Generator().manual_seed(42),
111
+ ).images[0]
112
+
113
+ image.save("output.png")
114
+ ```
115
+
116
+ 4. Launch WebUI and use the model as usual.
117
 
118
+ **Note**: Please make sure Git is installed and environment is properly configured on your machine.
119
 
120
  ---
121
 
 
123
 
124
  ## Parameters
125
 
 
126
  - CFG: 4 ~ 5
127
+ - Steps: 28 ~ 35
128
+ - Sampling Method: **Euler** (⚠️ Other samplers will not work properly)
129
+ - Resolution: Total area around 1024x1024. Best to choose from: 768x1344, **832x1216**, 896x1152, 1024x1024, 1152x896, 1216x832, 1344x768
130
 
131
  ## Prompts
132
+
133
+ - Prompt Prefix:
134
+
135
  ```
136
  masterpiece, best quality, newest, absurdres, highres, safe,
137
  ```
138
+
139
  - Negative Prompt:
140
+
141
  ```
142
+ nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro
143
  ```
144
 
145
  # Usage Guidelines
146
 
147
  ## Caption
148
+
149
+ ```
150
+ <1girl/1boy/1other/...>, <character>, <series>, <artists>, <special tags>, <general tags>, <other tags>
151
+ ```
152
 
153
  ## Quality Tags
154
+
155
+ For quality tags, we evaluated image popularity through the following process:
156
+
157
+ - Data normalization based on various sources and ratings.
158
+ - Application of time-based decay coefficients according to date recency.
159
+ - Ranking of images within the entire dataset based on this processing.
160
+
161
  Our ultimate goal is to ensure that quality tags effectively track user preferences in recent years.
162
 
163
+ | Percentile Range | Quality Tags |
164
+ | :--------------- | :------------- |
165
+ | > 95th | masterpiece |
166
+ | > 85th, <= 95th | best quality |
167
+ | > 60th, <= 85th | good quality |
168
+ | > 30th, <= 60th | normal quality |
169
+ | <= 30th | worst quality |
170
+
171
+ ## Aesthetic Tags
172
+
173
+ | Tag | Description |
174
+ | :-------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
175
+ | very awa | Top 5% of images in terms of aesthetic score by [waifu-scorer](https://huggingface.co/Eugeoter/waifu-scorer-v4-beta) |
176
+ | worst aesthetic | All the bottom 5% of images in terms of aesthetic score by [waifu-scorer](https://huggingface.co/Eugeoter/waifu-scorer-v4-beta) and [aesthetic-shadow-v2](https://huggingface.co/shadowlilac/aesthetic-shadow-v2) |
177
+ | ... | ... |
178
+
179
+ ## Date Tags
180
+
181
+ There are two types of date tags: **year tags** and **period tags**. For year tags, use `year xxxx` format, i.e., `year 2021`. For period tags, please refer to the following table:
182
+
183
+ | Year Range | Period tag |
184
+ | :--------- | :--------- |
185
+ | 2005-2010 | old |
186
+ | 2011-2014 | early |
187
+ | 2014-2017 | mid |
188
+ | 2018-2020 | recent |
189
+ | 2021-2024 | newest |
190
+
191
+ ## Dataset
192
+
193
+ - The latest Danbooru images up to the training date (approximately before 2024-10-23)
194
  - E621 images [e621-2024-webp-4Mpixel](https://huggingface.co/datasets/NebulaeWis/e621-2024-webp-4Mpixel) dataset on Hugging Face
195
 
196
  **Communication**
197
 
198
+ - **QQ Groups:**
 
 
 
199
 
200
+ - 875042008
201
+ - 914818692
202
+ - 635772191
203
 
204
+ - **Discord:** [Laxhar Dream Lab SDXL NOOB](https://discord.com/invite/DKnFjKEEvH)
205
 
206
  # Model License
207
 
208
+ This model's license inherits from https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0 fair-ai-public-license-1.0-sd and adds the following terms. Any use of this model and its variants is bound by this license.
209
 
210
  ## I. Usage Restrictions
211
 
 
233
 
234
  ## Participants
235
 
236
+ - **L_A_X:** [Civitai](https://civitai.com/user/L_A_X) | [Liblib.art](https://www.liblib.art/userpage/9e1b16538b9657f2a737e9c2c6ebfa69) | [Huggingface](https://huggingface.co/LAXMAYDAY)
237
+ - **li_li:** [Civitai](https://civitai.com/user/li_li) | [Huggingface](https://huggingface.co/heziiiii)
238
+ - **nebulae:** [Civitai](https://civitai.com/user/kitarz) | [Huggingface](https://huggingface.co/NebulaeWis)
239
+ - **Chenkin:** [Civitai](https://civitai.com/user/Chenkin) | [Huggingface](https://huggingface.co/windsingai)
240
+ - **Euge:** [Civitai](https://civitai.com/user/Euge_) | [Huggingface](https://huggingface.co/Eugeoter) | [Github](https://github.com/Eugeoter)
 
241
 
242
  ## Contributors
243
 
244
+ - **Narugo1992**: Thanks to [narugo1992](https://github.com/narugo1992) and the [deepghs](https://huggingface.co/deepghs) team for open-sourcing various training sets, image processing tools, and models.
245
+
246
+ - **Mikubill**: Thanks to [Mikubill](https://github.com/Mikubill) for the [Naifu](https://github.com/Mikubill/naifu) trainer.
247
+
248
+ - **Onommai**: Thanks to [OnommAI](https://onomaai.com/) for open-sourcing a powerful base model.
249
+
250
+ - **V-Prediction**: Thanks to the following individuals for their detailed instructions and experiments.
251
+
252
+ - adsfssdf
253
+ - [bluvoll](https://civitai.com/user/bluvoll)
254
+ - [bvhari](https://github.com/bvhari)
255
+ - [catboxanon](https://github.com/catboxanon)
256
+ - [parsee-mizuhashi](https://huggingface.co/parsee-mizuhashi)
257
+ - [very-aesthetic](https://github.com/very-aesthetic)
258
+
259
+ - **Community**: [aria1th261](https://civitai.com/user/aria1th261), [neggles](https://github.com/neggles/neurosis), [sdtana](https://huggingface.co/sdtana), [chewing](https://huggingface.co/chewing), [irldoggo](https://github.com/irldoggo), [reoe](https://huggingface.co/reoe), [kblueleaf](https://civitai.com/user/kblueleaf), [Yidhar](https://github.com/Yidhar), ageless, 白玲可, Creeper, KaerMorh, 吟游诗人, SeASnAkE, [zwh20081](https://civitai.com/user/zwh20081), Wenaka⁧~喵, 稀里哗啦, 幸运二副, 昨日の約, 445, [EBIX](https://civitai.com/user/EBIX), [Sopp](https://huggingface.co/goyishsoyish), [Y_X](https://civitai.com/user/Y_X), [Minthybasis](https://civitai.com/user/Minthybasis), [Rakosz](https://civitai.com/user/Rakosz)