Commit
•
c0165d8
1
Parent(s):
0556494
Update README.md
Browse files
README.md
CHANGED
@@ -51,16 +51,15 @@ the model avoids making mistakes that would make the generation more boring.
|
|
51 |
## Versions
|
52 |
|
53 |
### boring_e621
|
54 |
-
|
55 |
- **Description**: The first proof of concept of this idea. It is less refined than the other versions, but its less-intense effect might still be desirable.
|
56 |
- **Training Data Description**: A random selection of about 200 several-month-old images from e621.net having 0 favorites, fewer than 10(ish) up or down votes, and not having the tag "comic".
|
57 |
- **Model Trained On**: [YE18](https://huggingface.co/Doubleyobro/yiffy-e18).
|
58 |
- **Use Case**: Legacy. Maybe still useful on Stable Diffusion 1.4 or 1.5-based models if you want a less intense effect.
|
59 |
- **Trigger Word**: by boring_e621
|
|
|
60 |
|
61 |
|
62 |
### boring_e621_v4
|
63 |
-
|
64 |
- **Description**: Extends the basic Boring embedding idea with several improvements including:
|
65 |
- Tweaks to how data was selected for inclusion.
|
66 |
- Initialized with text "by <average vector representation of 100 low-scoring e621 artist names>" (which means that you should not have to use the word "by" to trigger it anymore)
|
@@ -68,24 +67,25 @@ the model avoids making mistakes that would make the generation more boring.
|
|
68 |
- **Model Trained On**: [YE18](https://huggingface.co/Doubleyobro/yiffy-e18).
|
69 |
- **Use Case**: The most general purpose embedding. Probably the best Boring embedding for most SD 1.4 or 1.5 based models.
|
70 |
- **Trigger Word**: boring_e621_v4
|
|
|
71 |
|
72 |
|
73 |
### boring_e621_fluffyrock_v4
|
74 |
-
|
75 |
- **Description**: A version of boring_e621_v4 trained on a different popular model that understands e621 tags.
|
76 |
- **Training Data Description**: Identical training set to boring_e621_v4.
|
77 |
- **Model Trained On**: [Some early non v-pred version of Fluffyrock](https://huggingface.co/lodestones/furryrock-model-safetensors).
|
78 |
- **Use Case**: Best for models having Fluffyrock in their ancestry, but might not work well on versions with "v-pred" in the name.
|
79 |
- **Trigger Word**: boring_e621_fluffyrock_v4
|
|
|
80 |
|
81 |
|
82 |
### boring_pony_v1
|
83 |
-
|
84 |
- **Description**: A proof of concept using a dataset collected from derpibooru.org and trained on a model that understands derpibooru tags.
|
85 |
- **Training Data Description**: About 200 several-month-old images from derpibooru.org having 0 favorites, a score of <= 0, and fewer than 5 total up or down votes.
|
86 |
- **Model Trained On**: [Pony Diffusion V4](https://huggingface.co/AstraliteHeart/pony-diffusion-v4)
|
87 |
- **Use Case**: None yet. It performs better on Pony Diffusion V4 than this model's in-built score-based negative-embedding-like tag (derpibooru_p_low), but less well than boring_e621_v4.
|
88 |
- **Trigger Word**: boring_pony_v1
|
|
|
89 |
|
90 |
|
91 |
# Bias, Risks, and Limitations
|
@@ -94,6 +94,7 @@ the model avoids making mistakes that would make the generation more boring.
|
|
94 |
* Unlike other negative embeddings, the Boring embeddings are not intended to fix problems like extra limbs or deformed hands. They can be used alongside other negative embeddings to fix deformities.
|
95 |
<br>
|
96 |
|
|
|
97 |
# Evaluation
|
98 |
|
99 |
To qualitatively illustrate how well the Boring embeddings have learned to improve image quality, we apply them to a small set of simple sample prompts using the base Stable Diffusion 1.5 model.
|
|
|
51 |
## Versions
|
52 |
|
53 |
### boring_e621
|
|
|
54 |
- **Description**: The first proof of concept of this idea. It is less refined than the other versions, but its less-intense effect might still be desirable.
|
55 |
- **Training Data Description**: A random selection of about 200 several-month-old images from e621.net having 0 favorites, fewer than 10(ish) up or down votes, and not having the tag "comic".
|
56 |
- **Model Trained On**: [YE18](https://huggingface.co/Doubleyobro/yiffy-e18).
|
57 |
- **Use Case**: Legacy. Maybe still useful on Stable Diffusion 1.4 or 1.5-based models if you want a less intense effect.
|
58 |
- **Trigger Word**: by boring_e621
|
59 |
+
<br>
|
60 |
|
61 |
|
62 |
### boring_e621_v4
|
|
|
63 |
- **Description**: Extends the basic Boring embedding idea with several improvements including:
|
64 |
- Tweaks to how data was selected for inclusion.
|
65 |
- Initialized with text "by <average vector representation of 100 low-scoring e621 artist names>" (which means that you should not have to use the word "by" to trigger it anymore)
|
|
|
67 |
- **Model Trained On**: [YE18](https://huggingface.co/Doubleyobro/yiffy-e18).
|
68 |
- **Use Case**: The most general purpose embedding. Probably the best Boring embedding for most SD 1.4 or 1.5 based models.
|
69 |
- **Trigger Word**: boring_e621_v4
|
70 |
+
<br>
|
71 |
|
72 |
|
73 |
### boring_e621_fluffyrock_v4
|
|
|
74 |
- **Description**: A version of boring_e621_v4 trained on a different popular model that understands e621 tags.
|
75 |
- **Training Data Description**: Identical training set to boring_e621_v4.
|
76 |
- **Model Trained On**: [Some early non v-pred version of Fluffyrock](https://huggingface.co/lodestones/furryrock-model-safetensors).
|
77 |
- **Use Case**: Best for models having Fluffyrock in their ancestry, but might not work well on versions with "v-pred" in the name.
|
78 |
- **Trigger Word**: boring_e621_fluffyrock_v4
|
79 |
+
<br>
|
80 |
|
81 |
|
82 |
### boring_pony_v1
|
|
|
83 |
- **Description**: A proof of concept using a dataset collected from derpibooru.org and trained on a model that understands derpibooru tags.
|
84 |
- **Training Data Description**: About 200 several-month-old images from derpibooru.org having 0 favorites, a score of <= 0, and fewer than 5 total up or down votes.
|
85 |
- **Model Trained On**: [Pony Diffusion V4](https://huggingface.co/AstraliteHeart/pony-diffusion-v4)
|
86 |
- **Use Case**: None yet. It performs better on Pony Diffusion V4 than this model's in-built score-based negative-embedding-like tag (derpibooru_p_low), but less well than boring_e621_v4.
|
87 |
- **Trigger Word**: boring_pony_v1
|
88 |
+
<br>
|
89 |
|
90 |
|
91 |
# Bias, Risks, and Limitations
|
|
|
94 |
* Unlike other negative embeddings, the Boring embeddings are not intended to fix problems like extra limbs or deformed hands. They can be used alongside other negative embeddings to fix deformities.
|
95 |
<br>
|
96 |
|
97 |
+
|
98 |
# Evaluation
|
99 |
|
100 |
To qualitatively illustrate how well the Boring embeddings have learned to improve image quality, we apply them to a small set of simple sample prompts using the base Stable Diffusion 1.5 model.
|