js-lee-AI / emotional-chatbot_gpt2

interactive and emotional chatbot

unsupervised-chatbot-GPT2

fine-tuning GPT-2 and Implement text generation chatbot This project aims to develop meorable and emotional chatbot using transfer learning (fine tune GPT-2 345M). You can find original code here.

It is never designed for commercial purposes.

Result

Install python library:

This project can be used regardless of tensorflow 1.x and tensorflow 2.x.

$ pip install tensorflow

$ pip install -r requirements.txt

Model install

1. clik the link and download.
2. Place the downloaded model in models\345M_org.

Usage

just run main.py

or

if you want to use your command line

$ python main.py

if you want to set hyperparameters

$ python main.py --top_k 10 --temperature 0.9 --nsamples 3

My dataset

My dataset is a .txt file (760 KB) of conversation between a bot and a user (my own file).

example is below

Author

Jungseob Lee / js-lee-AI / [e-mail]

Jungmu Park / boong_u / [e-mail]

Related papers

A Radford, et al., "Language Models are Unsupervised Multitask Learners", openAI blog, 2019.

A Vaswani, et al., "Attention is All you Need", NIPS 2017

Refrences

openAI

License

Modified MIT