|
|
--- |
|
|
license: mit |
|
|
datasets: |
|
|
- Adolfo-GM/AggmLite |
|
|
language: |
|
|
- en |
|
|
--- |
|
|
# AggmGPT |
|
|
|
|
|
AggmGPT is a lightweight foundational AI language model developed by **Adolfo GM**, designed to generate human-like text using n-gram models combined with self-attention mechanisms. The project is licensed under the **MIT License**, making it open-source and free for modification and distribution. |
|
|
|
|
|
## Features |
|
|
- Implements **Self-Attention** and **Multi-Head Attention** for improved context understanding. |
|
|
- Includes **positional encoding** to retain sequence structure. |
|
|
- Features a **Feed-Forward Neural Network** to refine predictions. |
|
|
- Tokenization and embedding functions for handling input text. |
|
|
|
|
|
## Installation |
|
|
Ensure you have Python installed on your system. You can download it from the [official website](https://www.python.org/downloads/). |
|
|
|
|
|
## How It Works |
|
|
AggmGPT processes text through: |
|
|
1. **Tokenization & Embedding** |
|
|
2. **Positional Encoding** |
|
|
3. **Self-Attention & Multi-Head Attention** |
|
|
4. **Feed-Forward Network** |
|
|
5. **N-gram Models** |
|
|
|
|
|
## License |
|
|
This project is licensed under the **MIT License**. See the `LICENSE` file for details. |
|
|
|
|
|
## Author |
|
|
Created by **Adolfo GM**. |
|
|
|
|
|
 |
|
|
 |
