Update README.md

README.md

@@ -2,16 +2,18 @@
 
 This repository contains the implementation of a lightweight, modified version of the GPT architecture **Leap-0** trained from scratch using FineWeb-Edu, an open-source dataset. The project demonstrates the design, training, and optimization of a custom natural language model on local hardware.  
 
+<div align="center">
+  <img src="LLM.drawio.png" alt="Description of the image" width="300">
+   <p><strong>Figure 1: Architecture of Leap</p>
+</div>
+
 ## Features  
 - **Custom GPT Architecture**: A miniaturized version of the GPT model tailored for efficient training on limited hardware.  
 - **Local Training**: Complete model training executed on local resources, enabling cost-effective development.  
 - **Open-Source Datasets**: Trained using publicly available FineWeb-Edu dataset to ensure accessibility and reproducibility.  
 - **Scalable Design**: Architecture optimized for experimentation and scalability while maintaining resource efficiency.  
 
-<div align="center">
-  <img src="LLM.drawio.png" alt="Description of the image" width="300">
-   <p><strong>Figure 1: Architecture of Leap</p>
-</div>
+
 
 ## Implementation Details  
 1. **Model Architecture**