A GTP-2 1.5b chatbot for Telegram with persistent sessions.
his project was made to test the GTP-2 1.5b model using Google Colaboratory and Telegram APIs. It works like a chatbot by tricking the model into completing a conversation and parsing the output. GTP2 1.5B replies could be made more accurate by fine-tuning the model using chats as a dataset; however, due to the GPU limitations of Google Colab and my lack of machine learning hardware I was not able to do so myself.
This project makes use of the following software:
- Node.js
- Telegram APIs (a BOT api key is needed)
- Google Colaboratory
- MongoDB Database
See the relative software installation documentation
https://github.com/McLavish/telegramBot_WebPanel
ESLint recommended coding style
- Luca P.
See also the list of contributors who participated in this project.
This project is licensed under the MIT License - see the LICENSE file for details