DerpGPT

A gloriously dumb AI chatbot that tries its best (but often fails).

🤖 What is DerpGPT?

DerpGPT is a conversational AI model that generates responses based on synthetic persona-chat datasets. It was trained on multi-turn dialogues but often struggles to keep things coherent. Whether it's making up bizarre responses, misunderstanding context, or just plain derping out, DerpGPT is here to entertain. You can downlaod him from my google drive: https://drive.google.com/file/d/16w5NxYWvVDqfaw9Hrojb1Xcw67ilBViv/view?usp=sharing

🔥 Features

• Word-Level Tokenization – Because subword tokenization would be too smart.
• Multi-Turn Dialogue Training – Trained on synthetic conversations, yet somehow still confused.
• Transformer-Based Architecture – A 50M parameter model doing its best.
• Blissful Cluelessness – Expect unexpected answers.

🎯 How It Works

1. The user sends a message.
2. DerpGPT attempts to generate a relevant response.
3. It either succeeds, produces nonsense, or accidentally continues the conversation forever.

🛠️ Technical Overview

• Model: Custom Transformer (GPT-style, ~50M parameters)
• Training Data: Synthetic multi-turn persona-chat dialogues
• Hardware: Runs on CUDA / MPS (if available)
• Training Framework: PyTorch

🚀 Usage

python derpgpt.py

Then just start chatting with DerpGPT and embrace the chaos.

🧠 Known Issues

• Sometimes generates entire conversations instead of single replies
• Occasionally forgets it’s an AI and makes weird claims
• Might just keep talking forever if not stopped

🤡 Why DerpGPT?

Because not every AI needs to be smart. Sometimes, you just need an AI that derps its way through conversations for the sake of entertainment.

🎤 Try it out and let the derping begin!