gpt/demos

Demos

This directory contains demonstration scripts that showcase the features of the GPT implementation.

Available Demos

• simple_demo.py: Simple demonstration of unlimited context with middle token eviction
• unlimited_context_demo.py: Comprehensive demo of the dynamic context management system
• version_demo.py: Demonstration of switching between different GPT versions

Running the Demos

Simple Demo

A basic demonstration of unlimited context with middle token eviction:

python demos/simple_demo.py

This demo shows how the system:

• Preserves beginning and end tokens
• Evicts middle tokens when memory constraints are reached
• Maintains context coherence despite token eviction

Unlimited Context Demo

A more comprehensive demonstration of the dynamic context system:

python demos/unlimited_context_demo.py

Features demonstrated:

• Token importance scoring
• Memory monitoring and management
• Dynamic token eviction strategies
• Preservation of semantic structure

Version Demo

Demonstration of loading and using different GPT versions:

python demos/version_demo.py

This shows:

• Configuration differences between versions
• Model parameter scaling
• Special features of each version
• Unlimited context configuration for GPT-5