AlekseyCalvin
commited on
Commit
•
c246c7a
1
Parent(s):
a311703
Update README.md
Browse files
README.md
CHANGED
@@ -5,101 +5,135 @@ tags:
|
|
5 |
- lora
|
6 |
- diffusers
|
7 |
- template:sd-lora
|
|
|
|
|
8 |
- ai-toolkit
|
9 |
widget:
|
10 |
-
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
output:
|
42 |
url: samples/1729701488064__000001800_0.jpg
|
43 |
-
- text:
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
|
|
46 |
output:
|
47 |
url: samples/1729711540947__000004000_0.jpg
|
48 |
-
- text:
|
49 |
-
|
50 |
-
|
|
|
|
|
|
|
|
|
51 |
output:
|
52 |
url: samples/1729700575417__000001600_0.jpg
|
53 |
-
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
98 |
base_model: stabilityai/stable-diffusion-3.5-large
|
99 |
-
|
100 |
license: creativeml-openrail-m
|
101 |
-
|
102 |
-
|
|
|
|
|
103 |
---
|
104 |
|
105 |
# HSTsd3ii
|
@@ -110,12 +144,14 @@ Model trained with [AI Toolkit by Ostris](https://github.com/ostris/ai-toolkit)
|
|
110 |
|
111 |
'HST style autochrome photo'
|
112 |
|
113 |
-
## Config Parameters
|
114 |
-
|
|
|
115 |
I've used A100 via Colab Pro.
|
116 |
-
|
117 |
-
|
118 |
-
|
|
|
119 |
```
|
120 |
job: extension
|
121 |
config:
|
@@ -165,7 +201,7 @@ config:
|
|
165 |
is_v3: true
|
166 |
quantize: true
|
167 |
```
|
168 |
-
Plus validation settings
|
169 |
|
170 |
## Download model and use it with ComfyUI, AUTOMATIC1111, SD.Next, Invoke AI, etc.
|
171 |
|
@@ -185,5 +221,4 @@ image = pipeline('HST style communist poster with text "JOIN RCA!", over autochr
|
|
185 |
image.save("my_image.png")
|
186 |
```
|
187 |
|
188 |
-
For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)
|
189 |
-
|
|
|
5 |
- lora
|
6 |
- diffusers
|
7 |
- template:sd-lora
|
8 |
+
- StableDiffusion3
|
9 |
+
- StableDiffusion3.5
|
10 |
- ai-toolkit
|
11 |
widget:
|
12 |
+
- text: >-
|
13 |
+
/@step 4000 weights:/ HST style autochrome photo of a drugged God
|
14 |
+
Christ-insect in wooden pillbox being pinned needle-poked by a giant Marina
|
15 |
+
Tsvetaeva in a long red gown, arcane symbolism, colorful HD realistic film
|
16 |
+
in Yury Annenkov & David Lynch styles, detailed face worrying, full-height
|
17 |
+
red-robed woman roams European city circa 1930
|
18 |
+
output:
|
19 |
+
url: samples/1729711624663__000004000_4.jpg
|
20 |
+
- text: >-
|
21 |
+
/@step 1600 weights:/HST style autochrome photo of a drugged God
|
22 |
+
Christ-insect in wooden pillbox being pinned needle-poked by a giant Marina
|
23 |
+
Tsvetaeva in a long red gown, arcane symbolism, colorful HD realistic film
|
24 |
+
in Yury Annenkov & David Lynch styles, detailed face worrying, full-height
|
25 |
+
red-robed woman roams European city circa 1930
|
26 |
+
output:
|
27 |
+
url: samples/1729700659095__000001600_4.jpg
|
28 |
+
- text: >-
|
29 |
+
/@step 1800 weights:/HST style autochrome photo of a drugged God
|
30 |
+
Christ-insect in wooden pillbox being pinned needle-poked by a giant Marina
|
31 |
+
Tsvetaeva in a long red gown, arcane symbolism, colorful HD realistic film
|
32 |
+
in Yury Annenkov & David Lynch styles, detailed face worrying, full-height
|
33 |
+
red-robed woman roams European city circa 1930
|
34 |
+
output:
|
35 |
+
url: samples/1729701571749__000001800_4.jpg
|
36 |
+
- text: >-
|
37 |
+
/@step 800 weights:/HHST style autochrome photo of a drugged God
|
38 |
+
Christ-insect in wooden pillbox being pinned needle-poked by a giant Marina
|
39 |
+
Tsvetaeva in a long red gown, arcane symbolism, colorful HD realistic film
|
40 |
+
in Yury Annenkov & David Lynch styles, detailed face worrying, full-height
|
41 |
+
red-robed woman roams European city circa 1930
|
42 |
+
output:
|
43 |
+
url: samples/1729696985671__000000800_4.jpg
|
44 |
+
- text: >-
|
45 |
+
/@step 1800 weights:/HST style communist poster with text "JOIN RCA!", over
|
46 |
+
autochrome color photo of Vladimir Lenin at a Dada cabaret in 1916 Zurich,
|
47 |
+
dancing with red feathered drunken dinosaur, an early conceptual artist.
|
48 |
+
Lenin is full of contageous awe, his blemished skin flushing with anxious
|
49 |
+
excitement, his famous bald spot sweatily glistening under warm lights. In
|
50 |
+
the back, Krupskaya and Inessa Armand laugh.
|
51 |
output:
|
52 |
url: samples/1729701488064__000001800_0.jpg
|
53 |
+
- text: >-
|
54 |
+
/@step 4000 weights:/HHST style communist poster with text "JOIN RCA!", over
|
55 |
+
autochrome color photo of Vladimir Lenin at a Dada cabaret in 1916 Zurich,
|
56 |
+
dancing with red feathered drunken dinosaur, an early conceptual artist.
|
57 |
+
Lenin is full of contageous awe, his blemished skin flushing with anxious
|
58 |
+
excitement, his famous bald spot sweatily glistening under warm lights. In
|
59 |
+
the back, Krupskaya and Inessa Armand laugh.
|
60 |
output:
|
61 |
url: samples/1729711540947__000004000_0.jpg
|
62 |
+
- text: >-
|
63 |
+
/@step 1600 weights:/ HST style communist poster with text "JOIN RCA!", over
|
64 |
+
autochrome color photo of Vladimir Lenin at a Dada cabaret in 1916 Zurich,
|
65 |
+
dancing with red feathered drunken dinosaur, an early conceptual artist.
|
66 |
+
Lenin is full of contageous awe, his blemished skin flushing with anxious
|
67 |
+
excitement, his famous bald spot sweatily glistening under warm lights. In
|
68 |
+
the back, Krupskaya and Inessa Armand laugh.
|
69 |
output:
|
70 |
url: samples/1729700575417__000001600_0.jpg
|
71 |
+
- text: >-
|
72 |
+
/@step 1600 weights:/ HST autochrome style analog dslr award-winning 8k art
|
73 |
+
photo showing a nurse battling a giant cell phone, above text caption of
|
74 |
+
\ONCE WE WERE ALL HARRY GARDNER!\, in a highly realistic modern American
|
75 |
+
medical hospital
|
76 |
+
output:
|
77 |
+
url: samples/1729700596338__000001600_1.jpg
|
78 |
+
- text: >-
|
79 |
+
/@step 1000 weights:/ HST autochrome style analog dslr award-winning 8k art
|
80 |
+
photo showing a nurse battling a giant cell phone, above text caption of
|
81 |
+
\ONCE WE WERE ALL HARRY GARDNER!\, in a highly realistic modern American
|
82 |
+
medical hospital
|
83 |
+
output:
|
84 |
+
url: samples/1729697845170__000001000_1.jpg
|
85 |
+
- text: >-
|
86 |
+
/@step 3600 weights:/ HST style autochrome photograph of a dark CIA agent
|
87 |
+
Koala leaping at the throat of goth freedom-fighter Julian Assange, tragic
|
88 |
+
atmosphere, award-winning art photo
|
89 |
+
output:
|
90 |
+
url: samples/1729709764758__000003600_2.jpg
|
91 |
+
- text: >-
|
92 |
+
/@step 1200 weights:/ HST style autochrome photograph of a dark CIA agent
|
93 |
+
Koala leaping at the throat of goth freedom-fighter Julian Assange, tragic
|
94 |
+
atmosphere, award-winning art photo
|
95 |
+
output:
|
96 |
+
url: samples/1729698783304__000001200_2.jpg
|
97 |
+
- text: >-
|
98 |
+
/@step 1000 weights:/ HST style autochrome photograph of a dark CIA agent
|
99 |
+
Koala leaping at the throat of goth freedom-fighter Julian Assange, tragic
|
100 |
+
atmosphere, award-winning art photo
|
101 |
+
output:
|
102 |
+
url: samples/1729697866093__000001000_2.jpg
|
103 |
+
- text: >-
|
104 |
+
/@step 4000 weights:/ HST style autochrome photo of realistic green-eyed
|
105 |
+
black cat, with prominent regions of white fur, playing a piano and singing,
|
106 |
+
amateur 2004 photograph shot on a cell phone in a Los Angeles apartment
|
107 |
+
kitchen
|
108 |
+
output:
|
109 |
+
url: samples/1729711603742__000004000_3.jpg
|
110 |
+
- text: >-
|
111 |
+
/@step 3000 weights:/ HST style autochrome photo of realistic green-eyed
|
112 |
+
black cat, with prominent regions of white fur, playing a piano and singing,
|
113 |
+
amateur 2004 photograph shot on a cell phone in a Los Angeles apartment
|
114 |
+
kitchen
|
115 |
+
output:
|
116 |
+
url: samples/1729707029633__000003000_3.jpg
|
117 |
+
- text: >-
|
118 |
+
/@step 1800 weights:/ HST style autochrome photo of realistic green-eyed
|
119 |
+
black cat, with prominent regions of white fur, playing a piano and singing,
|
120 |
+
amateur 2004 photograph shot on a cell phone in a Los Angeles apartment
|
121 |
+
kitchen
|
122 |
+
output:
|
123 |
+
url: samples/1729701550832__000001800_3.jpg
|
124 |
+
- text: >-
|
125 |
+
/@step 1000 weights:/ HST style autochrome photo of realistic green-eyed
|
126 |
+
black cat, with prominent regions of white fur, playing a piano and singing,
|
127 |
+
amateur 2004 photograph shot on a cell phone in a Los Angeles apartment
|
128 |
+
kitchen
|
129 |
+
output:
|
130 |
+
url: samples/1729697887015__000001000_3.jpg
|
131 |
base_model: stabilityai/stable-diffusion-3.5-large
|
|
|
132 |
license: creativeml-openrail-m
|
133 |
+
language:
|
134 |
+
- en
|
135 |
+
pipeline_tag: text-to-image
|
136 |
+
library_name: diffusers
|
137 |
---
|
138 |
|
139 |
# HSTsd3ii
|
|
|
144 |
|
145 |
'HST style autochrome photo'
|
146 |
|
147 |
+
## Config Parameters
|
148 |
+
*Dim:16 Alpha:32 Optimizer:Ademamix8bit LR:2e-4* **More info below!** <br>
|
149 |
+
Fine-tuned using the **Google Colab Notebook*** Version of **ai-toolkit**.<br>
|
150 |
I've used A100 via Colab Pro.
|
151 |
+
However, training SD3.5 may potentially work with Free Colab or lower Vram in general:<br>
|
152 |
+
Especially if one were to use lower rank (try 4 or 8), dataset size (in terms of caching/bucketing/pre-loading impacts), 1 batch size, Adamw8bit optimizer, 512 resolution, maybe adding the "lowvram, true" argument, and plausibly specifying alternate quantization variants!
|
153 |
+
Generally, VRAM expenditures tend to be lower than for Flux during training. So, try it! I certainly will.<br>
|
154 |
+
**To use on Colab**, modify a Flux template Notebook from [here](https://github.com/ostris/ai-toolkit/tree/main/notebooks) with parameters from Ostris' example config for SD3.5 [here](https://github.com/ostris/ai-toolkit/blob/main/config/examples/train_lora_sd35_large_24gb.yaml)!
|
155 |
```
|
156 |
job: extension
|
157 |
config:
|
|
|
201 |
is_v3: true
|
202 |
quantize: true
|
203 |
```
|
204 |
+
Plus validation settings:<br> Prompts like the above, at 1024, guidance scale 4, 25 steps, seed 42, no negatives.
|
205 |
|
206 |
## Download model and use it with ComfyUI, AUTOMATIC1111, SD.Next, Invoke AI, etc.
|
207 |
|
|
|
221 |
image.save("my_image.png")
|
222 |
```
|
223 |
|
224 |
+
For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)
|
|