Note, this repo is deprecated.

If you are interested in further enhancements and investigations, just watch Next repo.

https://github.com/Samurais/Neural_Conversation_Models

Approaching a Chatbot Service

Part 1: Introduction

Part 2: Bot Engine

Part 3: Bot Model

This repository is align with Part 3: Bot Model.

Train and serve QA Model with TensorFlow

Tested with TensorFlow#0.11.0rc2, Python#3.5.

Install Nvidia Drivers, CUDNn, Python, TensorFlow on Ubuntu 16.04

DeepQA

Inspired and inherited from DeepQA.

Install deps

pip install -r requirements.txt

Install TensorFlow

export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow-0.11.0rc2-cp35-cp35m-linux_x86_64.whl
pip install —-upgrade $TF_BINARY_URL

Pre-process data

Process data, build vocabulary, word embedding, conversations, etc.

cp config.sample.ini config.ini
python deepqa2/dataset/preprocesser.py

Sample Corpus http://www.cs.cornell.edu/~cristian/Cornell_Movie-Dialogs_Corpus.html

Train Model

Train language model with Seq2seq.

cp config.sample.ini config.ini # modify keys
python deepqa2/train.py

Serve Model

Provide RESt API to access language model.

cd DeepQA2/save/deeplearning.cobra.vulcan.20170127.175256/deepqa2/serve
cp db.sample.sqlite3 db.sqlite3 
python manage.py runserver 0.0.0.0:8000

Access Service with RESt API

POST /api/v1/question HTTP/1.1
Host: 127.0.0.1:8000
Content-Type: application/json
Authorization: Basic YWRtaW46cGFzc3dvcmQxMjM=
Cache-Control: no-cache

{"message": "good to know"}

response
{
  "rc": 0,
  "msg": "hello"
}

Train with Docker

Install

• docker
• nvidia-docker

Train

docker pull samurais/deepqa2:latest
cd DeepQA2
./scripts/train_with_docker.sh