Skip to content
Giant Language Model Test Room
TypeScript Python HTML CSS JavaScript
Branch: master
Clone or download
Latest commit 88a04ba Apr 21, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
backend init. Apr 12, 2019
client init. Apr 12, 2019
figs init. Apr 12, 2019
.gitignore init. Apr 12, 2019
LICENSE Create LICENSE Apr 12, 2019
README.md Update README.md Apr 21, 2019
requirements.txt init. Apr 12, 2019
server.py init. Apr 12, 2019
server.yaml init. Apr 12, 2019

README.md

GLTR: Giant Language Model Test Room

Detecting text that was generated from large language models (e.g. GPT-2).

webpage: http://gltr.io
online-demo: http://gltr.io/dist/index.html
paper: PENDING

A project by Hendrik Strobelt, Sebastian Gehrmann, Alexander M. Rush.

collaboration of MIT-IBM Watson AI Lab and HarvardNLP

Quickstart

Install dependencies for Python >3.6 :

pip install -r requirements.txt

run server for gpt-2-small:

python server.py

the demo instance runs now at http://localhost:5001/client/index.html

Run the BERT server

start the server for BERT:

python server.py --model BERT

the instance runs now at http://localhost:5001/client/index.html?nodemo. HINT: we only provide demo texts for gpt2-small.

server.py options

usage: server.py [-h] [--model MODEL] [--nodebug NODEBUG] [--address ADDRESS]
                 [--port PORT] [--nocache NOCACHE] [--dir DIR] [--no_cors]

optional arguments:
  -h, --help         show this help message and exit
  --model MODEL		 choose either 'gpt-2-small' (default) or 'BERT' or your own
  --nodebug NODEBUG  server in non-debugging mode
  --port PORT	     port to launch UI and API (default:5001)
  --no_cors          launch API without CORS support (default: False)

Extend backend

The backend defines a number of model api's that can be invoked by the server by starting it with the parameter --model NAME. To add a custom model, you need to write your own api in backend/api.py and add the decorator @register_api(name=NAME).

Each api needs to be a class that inherits from AbstractLanguageChecker, which defines two functions check_probabilities and postprocess. Please follow the documentation within api.py when implementing the class and the functions.

Extend frontend

the source code for the front-end is in client/src.

To modify, installing of node dependencies is necessary:

cd client/src; npm install; cd ../..

re-compilation of front-end:

> rm -rf client/dist;cd client/src/; npm run build; cd ../..

License

Apache 2

(c) 2019 by Hendrik Strobelt, Sebastian Gehrmann, Alexander M. Rush

You can’t perform that action at this time.