Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,58 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- ar
|
5 |
+
pipeline_tag: text-generation
|
6 |
+
tags:
|
7 |
+
- 'arabic '
|
8 |
+
- text-generation
|
9 |
---
|
10 |
+
# ArabianGPT Model Overview
|
11 |
+
|
12 |
+
## Introduction
|
13 |
+
ArabianGPT-0.3B, developed under the ArabianLLM initiatives, is a specialized GPT-2 model optimized for Arabic language modeling.
|
14 |
+
It's a product of the collaborative efforts at Prince Sultan University's Robotics and Internet of Things Lab, focusing on enhancing natural language modeling and generation in Arabic.
|
15 |
+
This model represents a significant stride in LLM research, specifically addressing the linguistic complexities and nuances of the Arabic language.
|
16 |
+
|
17 |
+
## Key Features
|
18 |
+
- **Architecture**: GPT-2
|
19 |
+
- **Model Size**: 345 million parameters
|
20 |
+
- **Layers**: 24
|
21 |
+
- **Model Attention Layers (MAL)**: 16
|
22 |
+
- **Context Window Size**: 1024 tokens
|
23 |
+
|
24 |
+
## Training
|
25 |
+
- **Dataset**: C4, Twitter, Wiki
|
26 |
+
- **Data Size**: 23 GB
|
27 |
+
- **Tokenizer**: Aranizer 64K
|
28 |
+
- **Tokens**: Over 3.3 billion
|
29 |
+
- **Hardware**: 4 NDIVIA A100 GPUs
|
30 |
+
- **Training Duration**: 45 days
|
31 |
+
- **Performance**: loss of 3.88
|
32 |
+
|
33 |
+
|
34 |
+
## Role in ArabianLLM Initiatives
|
35 |
+
ArabianGPT-0.3B is crucial for advancing Arabic language processing, addressing challenges unique to Arabic morphology and dialects.
|
36 |
+
|
37 |
+
## Usage
|
38 |
+
Suitable for Arabic text generation tasks. Example usage with Transformers Pipeline:
|
39 |
+
```python
|
40 |
+
from transformers import pipeline
|
41 |
+
|
42 |
+
pipe = pipeline("text-generation", model="riotu-lab/ArabianGPT-03B", max_new_tokens=512)
|
43 |
+
text = ''
|
44 |
+
pipe.predict(text)
|
45 |
+
```
|
46 |
+
|
47 |
+
## Limitations and Ethical Considerations
|
48 |
+
|
49 |
+
- The model may have context understanding or text generation limitations in certain scenarios.
|
50 |
+
- Emphasis on ethical use to prevent misinformation or harmful content propagation.
|
51 |
+
|
52 |
+
## Acknowledgments
|
53 |
+
|
54 |
+
Special thanks to Prince Sultan University, particularly the Robotics and Internet of Things Lab.
|
55 |
+
|
56 |
+
## Contact Information
|
57 |
+
|
58 |
+
For inquiries: [riotu@psu.edu.sa](mailto:riotu@psu.edu.sa).
|