Spaces:
Running
Running
Update README.md
Browse files
README.md
CHANGED
|
@@ -6,7 +6,6 @@ colorTo: red
|
|
| 6 |
sdk: static
|
| 7 |
pinned: false
|
| 8 |
---
|
| 9 |
-
|
| 10 |
# 🔍 OverseerAI
|
| 11 |
|
| 12 |
## Mission
|
|
@@ -17,22 +16,25 @@ OverseerAI is dedicated to advancing open-source AI safety and content moderatio
|
|
| 17 |
### Datasets
|
| 18 |
#### [BrandSafe-16k](https://huggingface.co/datasets/OverseerAI/BrandSafe-16k)
|
| 19 |
A comprehensive dataset for training brand safety classification models, featuring 16 distinct risk categories:
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
### Models
|
| 38 |
|
|
@@ -49,4 +51,31 @@ A lightweight, optimized version of vision-1:
|
|
| 49 |
- Architecture: Llama 3.1 8B
|
| 50 |
- Quantization: GGUF V3 (Q4_K)
|
| 51 |
- Optimized for Apple Silicon
|
| 52 |
-
- Fast load time:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
sdk: static
|
| 7 |
pinned: false
|
| 8 |
---
|
|
|
|
| 9 |
# 🔍 OverseerAI
|
| 10 |
|
| 11 |
## Mission
|
|
|
|
| 16 |
### Datasets
|
| 17 |
#### [BrandSafe-16k](https://huggingface.co/datasets/OverseerAI/BrandSafe-16k)
|
| 18 |
A comprehensive dataset for training brand safety classification models, featuring 16 distinct risk categories:
|
| 19 |
+
|
| 20 |
+
| Category | Description |
|
| 21 |
+
|----------|-------------|
|
| 22 |
+
| B1-PROFANITY | Explicit language and cursing |
|
| 23 |
+
| B2-OFFENSIVE_SLANG | Informal offensive terms |
|
| 24 |
+
| B3-COMPETITOR | Competitive brand mentions |
|
| 25 |
+
| B4-BRAND_CRITICISM | Negative brand commentary |
|
| 26 |
+
| B5-MISLEADING | Deceptive or false information |
|
| 27 |
+
| B6-POLITICAL | Political content and discussions |
|
| 28 |
+
| B7-RELIGIOUS | Religious themes and references |
|
| 29 |
+
| B8-CONTROVERSIAL | Contentious topics |
|
| 30 |
+
| B9-ADULT | Adult or mature content |
|
| 31 |
+
| B10-VIOLENCE | Violent themes or descriptions |
|
| 32 |
+
| B11-SUBSTANCE | Drug and alcohol references |
|
| 33 |
+
| B12-HATE | Hate speech and discrimination |
|
| 34 |
+
| B13-STEREOTYPE | Stereotypical content |
|
| 35 |
+
| B14-BIAS | Biased viewpoints |
|
| 36 |
+
| B15-UNPROFESSIONAL | Unprofessional content |
|
| 37 |
+
| B16-MANIPULATION | Manipulative content |
|
| 38 |
|
| 39 |
### Models
|
| 40 |
|
|
|
|
| 51 |
- Architecture: Llama 3.1 8B
|
| 52 |
- Quantization: GGUF V3 (Q4_K)
|
| 53 |
- Optimized for Apple Silicon
|
| 54 |
+
- Fast load time: 3.27s
|
| 55 |
+
- Efficient memory usage: 4552.80 MiB CPU / 132.50 MiB Metal
|
| 56 |
+
- Perfect for local deployment and smaller compute resources
|
| 57 |
+
|
| 58 |
+
## 💡 Use Cases
|
| 59 |
+
- Content moderation for social media platforms
|
| 60 |
+
- Brand safety monitoring for advertising
|
| 61 |
+
- User-generated content filtering
|
| 62 |
+
- Real-time content classification
|
| 63 |
+
- Safe content recommendation systems
|
| 64 |
+
|
| 65 |
+
## 🤝 Contributing
|
| 66 |
+
We welcome contributions from the community! Whether it's:
|
| 67 |
+
- Improving model accuracy
|
| 68 |
+
- Expanding the dataset
|
| 69 |
+
- Optimizing for different hardware
|
| 70 |
+
- Adding new classification categories
|
| 71 |
+
- Reporting issues or suggesting improvements
|
| 72 |
+
|
| 73 |
+
## 📫 Contact
|
| 74 |
+
- GitHub: [OverseerAI](https://github.com/OverseerAI)
|
| 75 |
+
- HuggingFace: [OverseerAI](https://huggingface.co/OverseerAI)
|
| 76 |
+
|
| 77 |
+
## 📜 License
|
| 78 |
+
Our models are released under the Llama 3.1 license, and our datasets are available under open-source licenses to promote accessibility and innovation in AI safety.
|
| 79 |
+
|
| 80 |
+
---
|
| 81 |
+
*OverseerAI - Making AI Safety Accessible and Efficient*
|