Skip to content
This repository has been archived by the owner on Jun 1, 2022. It is now read-only.

Latest commit

 

History

History
149 lines (107 loc) · 3.62 KB

README.md

File metadata and controls

149 lines (107 loc) · 3.62 KB
Raw

GPT-2 Test

Intention

To investigate the generated content provided by the GPT-2 117M learned model - GPT-2 repo.

Feature creep

  • Download each iteration as .txt
  • Download all iterations as .txt
  • On live, create request and output logs

Setup

  • Clone this repo
  • Open terminal and cd to the cloned repo location
brew install python
pip3 install virtualenv
cd src
virtualenv venv
source venv/bin/activate
pip3 install Flask flask-cors tensorflow==1.13.1
pip3 install -r requirements.txt
python3 download_model.py 117M
python3 download_model.py 345M
python3 download_model.py 774M

Running

In terminal:

python3 application.py
  • Go to http://0.0.0.0:80
  • Insert the text you want to use as a seed in the text field and hit submit
  • A waiting for repsonse... message will appear until the generated content is...uh...generated, then the content will replace the message

GPT-2 repo README

This repo is based on the original GPT-2 repo, you can read the setup instructions in this repo's GPT2-README.md

Creating the live server

Tried to get this up on Heroku or a Lambda but both have storage size limits which are too small for the 500MB model + repo code and requirements.

  • Heroku - 500MB
  • Lambda - 250MB

So had to settle for a custom EC2 build.

AWS EC2 instance

  • Step 1
    • Amazon Linux 2 AMI
  • Step 2
    • Select t2.medium
  • Step 3:
    • Auto-assign Public IP to Enable
    • Otherwise Default settings
  • Step 4
    • Set 12 GBs of storage
  • Step 5
  • Step 6
    • Use ftlabs-gpt2SG security gruop which grants:
      • SSH - FT internal IPs
      • HTTP - 0.0.0.0/0, ::/0
      • HTTPS - 0.0.0.0/0, ::/0
  • Step 7
    • Create using new key (currently ftlabs-gpt.pem) or ley you already have access to

Connect to instance

  • cd .aws
  • ssh -i "ftlabs-gpt2-key.pem" ec2-54-194-234-190.eu-west-1.compute.amazonaws.com

Installing the repo and requirements

sudo yum install python3 git gcc python3-devel
git clone https://github.com/ftlabs/gpt2-gui.git
cd ~/gpt2-gui/src
sudo pip3 install Flask flask-cors tensorflow==1.13.1
sudo pip3 install -r requirements.txt
sudo python3 download_model.py 117M
sudo python3 download_model.py 345M
sudo python3 download_model.py 774M

Add a cron job to restart the server on reboot

  • crontab -e
  • i
  • paste this line: screen -dmS gpt2-gui-server bash -c 'cd ~/gpt2-gui/src; sudo python3 application.py; exec sh'
  • press esc
  • type :wq

Running

  • Use screen -r to see running script (should be one from booting up server)
  • if not:
cd ~/gpt2-gui/src
screen -dmS gpt2-gui-server bash -c 'cd ~/gpt2-gui/src; sudo python3 application.py; exec sh'

cmd + shift + a then d

Updating

  • ssh into server
  • screen -r gpt2-gui-server
  • cmd + shift + a then d
  • y
cd gpt2-gui
git pull origin master
cd src
screen -dmS gpt2-gui-server bash -c 'cd ~/gpt2-gui/src; sudo python3 application.py; exec sh'

Server is now running and you can exit the ssh session.

Checking

  • screen -ls - see current running screens
  • screen -r <name> - access a specific screen

Close an exisitng screen

  • screen -r <name> - access a specific screen
  • cmd + shift + a then k
  • y

Appendix