🧠 AI Brain Explorer (Transformer Visualizer)

An interactive Streamlit application that demystifies how a Transformer processes sequences, computes attention, and generates contextual representations.

---

🌟 Overview

AI Brain Explorer is an educational and intuitive tool designed to visually break down the internal mechanics of Transformer models. From tokenization to layer-wise representations, this app allows you to explore exactly how Artificial Intelligence reads, understands, and forms connections between words.

✨ Features

The application is logically divided into 7 core visual components, each representing a crucial step in the Transformer architecture:

1. 📝 Word Splitting (Tokenization & Embeddings): See how sentences are split into individual tokens and converted into numerical "Meaning Vectors".
2. 📐 Word Order (Positional Encoding): Understand how the model distinguishes word sequences using unique sine/cosine mathematical wave patterns.
3. 🔗 Word Connections (Self-Attention): Dive into Query, Key, and Value vectors. Visualize raw attention scores and explore the interactive heatmap of softmax attention weights.
4. 🧠 Finding Patterns (Multi-Head Attention): Compare multiple attention heads in parallel and see how different heads focus on different syntactic or semantic relationships.
5. ⚡ Processing Meaning (Feed-Forward Network): Explore the position-wise FFN and see how the ReLU activation function introduces non-linearity.
6. ♻️ Brain Polish (Residuals & Layer Norm): Visualize network stabilization. See how skip connections prevent vanishing gradients, and how LayerNorm centers vector distributions.
7. 📊 Deep Understanding (Layer Stacking): Track token vectors across multiple consecutive layers. Observe how vector representations become increasingly contextual via Cosine Similarity and L2 Norm drop-offs.

🚀 Getting Started

Prerequisites

You need Python installed on your machine. Install the required dependencies using pip:

pip install streamlit numpy pandas plotly

Running the App

1. Navigate to the project directory:

cd Streamlint-Demo

2. Run the Streamlit server:

streamlit run app.py

3. The application will launch automatically in your default internet browser at http://localhost:8501.

⚙️ Interactive Sidebar

• ✏️ Input Sentence: Try out different sentences to see how the attention networks and embeddings react to changing vocabulary.
• ⚙️ Hyperparameters: Adjust the Number of Layers, Attention Heads, and Embedding Dimension on the fly to see how it affects the matrices.
• 🔘 Display Toggles: Toggle the visibility of complex graphs like Positional Encodings and Attention Maps.

🛠️ Tech Stack

• Streamlit: For the highly interactive, responsive web UI and custom CSS styling.
• Plotly Express & Graph Objects: For generating beautiful, scalable dark-themed heatmaps, bar charts, and histograms.
• NumPy & Pandas: For constructing the underlying matrix multiplications, linear algebra operations, and data logic safely from scratch.