Thus begins the chronicles of Dooracle; a fledgling digital product startup set to revolutionize the fortune telling industry.

Their moonshot?

To build a fortune-telling chatbot.

The project is aimed at filling the demand for people who want their futures predicted.

When the superintelligent AI is finally ready for transfer learning, their product will be there.

A bulletproof plan.

Step by Step NLP: Data to Deployment

A workshop on how to build and deploy a virtual chatbot using Natural Language Processing. You can check out the slide here

📹👇

After the event, we heard your feedback! Greg created a step by step walkthrough, complete with stumbles and rewinds, that takes about 75 minutes to see everything about what happens behind the scenes.

How To Use These Videos

The following videos were part of original workshop, and will help you to deploy the DialoGPT-medium chatbot to your local machine or to the cloud.

Prerequisites

These videos were shot on a Windows machine running Ubuntu on WSL2 through Windows Terminal. Google Cloud Platform was used to deploy the chatbot remotely. Docker Desktop was also leveraged along with Notepadd++

Part 1: Get Chatbot Running on Virtual Machine in Jupyter Notebook

Install packages in WSL2 Ubuntu Root

sudo apt update
sudo apt install python3-pip python3-dev
sudo -H pip3 install --upgrade pip
sudo -H pip3 install virtualenv

Allocate folders and create vm

mkdir ~/chatbot-event
cd chatbot-event
virtualenv chattyvm
source chattyvm/bin/activate

Install libraries

pip install jupyter
pip install torch
pip install transformers

Then just run DialoGPT in Jupyter!

Part 2: Open-Source Chatbot UI and DialoGPT Intelligence Upgrade

Install git and clone this repo!

sudo apt install git
git clone https://github.com/FourthBrain/step-by-step-nlp-dialogpt-chatbot.git

Run it!

cd step-by-step-nlp-dialogpt-chatbot/dialogpt-chatbot
flask run

Part 3: Dockerizing DialoGPT Chatbot

Install git and clone this repo!

sudo apt install git
git clone https://github.com/FourthBrain/step-by-step-nlp-dialogpt-chatbot.git

Run it!

cd ..
cd dialogpt-chatbot-dockerize
docker build -t dialogpt-chatbot-image
docker run -d -p 8080:8080 dialogpt-chatbot-image

Part 4: Deploying DialoGPT in the Cloud

Create account, setup billing, open terminal, upload files

cd dialogpt-chatbot-dockerize
gcloud builds submit --tag gcr.io/my-project/dialogpt-chatbot-image
gcloud run deploy --image gcr.io/my-project/dialogpt-chatbot-image --memory 4Gi

Shoutouts

Big ups to Hugging Face for their awesome repo of models and streamlining all things SOTA ML modeling; you can check out DialoGPT-medium here

Thanks to bhargavramdr for the right chatbot UI at the right time. Happy Halloween 2021!

License

All source from FourthBrain is free to use, but microsoft/DialoGPT is licensed under the MIT License.