77 lines (49 loc) · 2.31 KB

Distributed Chatbot Training over P2P

Train your own AI chatbot with friends — no data sharing, just local training and model merging over the network.

What is this?

Each client trains a Transformer chatbot on a private dataset.
The server sends data slices, collects models & vocab files.
It aggregates all models into a single smart brain 🧠

Run in 3 Steps

Start the Server

python server.py

Start One or More Clients

python train.py

Each client trains locally and sends back a model.

Chat with the AI

python runner.py

Ask anything, get smart replies. Internet features included.

Requirements

torch
pandas
requests
beautifulsoup4
serpapi

Install:

pip install -r requirements.txt

Cool Features

Transformer-based chatbot (PyTorch)
Weighted model merging (like federated learning)
Web integration: weather, news, search
Beam Search decoding
Auto-logs chat to CSV

License

MIT — use freely and improve it!

Like it?

Star the repo and tell your friends!