Skip to content
A Marvelous ChatBot implement using PyTorch.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
config
datasets first commit May 10, 2017
models first commit May 10, 2017
utils
.gitignore first commit May 10, 2017
README.md Update README.md Jun 10, 2019
train.py

README.md

PyTorch Marvelous ChatBot

[Update]

it's 2019 now, previously model can not catch up state-of-art now. So we just move towards the future a transformer based chatbot, it's much more accurate and flexiable as well as full of imagination interms of being a cahtbot!!! More importantly we are opensourced the whole codes here: http://manaai.cn/aicodes_detail3.html?id=36 Be sure to check it if you interested in chatbot and NLP!! it's build with tensorflow 2.0 newest api!!

Aim to build a Marvelous ChatBot

Synopsis

This is the first and the only opensource of ChatBot, I will continues update this repo personally, aim to build a intelligent ChatBot, as the next version of Jarvis.

This repo will maintain to build a Marvelous ChatBot based on PyTorch, welcome star and submit PR.

Already Done

Currently this repo did those work:

  • based on official tutorial, this repo will move on develop a seq2seq chatbot, QA system;
  • re-constructed whole project, separate mess code into data, model, train logic;
  • model can be save into local, and reload from previous saved dir, which is lack in official tutorial;
  • just replace the dataset you can train your own data!

Last but not least, this project will maintain or move on other repo in the future but we will continue implement a practical seq2seq based project to build anything you want: Translate Machine, ChatBot, QA System... anything you want.

Requirements

PyTorch
python3
Ubuntu Any Version
Both CPU and GPU can works

Usage

Before dive into this repo, you want to glance the whole structure, we have these setups:

  • config: contains the config params, which is global in this project, you can change a global param here;
  • datasets: contains data and data_loader, using your own dataset, you should implement your own data_loader but just a liitle change on this one;
  • models: contains seq2seq model definition;
  • utils: this folder is very helpful, it contains some code may help you get out of anoying things, such as save model, or catch KeyboardInterrupt exception or load previous model, all can be done in here.

to train model is also straightforward, just type:

python3 train.py

Contribute

wecome submit PR!!!! Let's build ChatBot together!

Contact

if you have anyquestion, you can find me via wechat jintianiloveu

You can’t perform that action at this time.