Flask-Joey: Deploy your Joey NMT model as a flask app

I created this app to serve our English to Fongbe translation model, but you can use it as a template to serve your model too without much coding. The model used was created using this notebook from the Masakhane project.

Setup

Clone this repository

git clone https://github.com/kevindegila/flask-joey

Install

cd into the repo and Create a python 3.7 virtual environment with your desired env manager. Here is an exemple with conda:

cd flask-joey && conda create --name translator python=3.7

Activate the environment and install all the requirements:

conda activate translator && pip install -r requirements.txt

Create a database

Run Python at your terminal

python3

Inside Python run the following command

>>> import sqlite3,os
>>> conn = sqlite3.connect("trReviews.sqlite")
>>> c = conn.cursor()
>>> c.execute('DROP TABLE IF EXISTS trReviews')
<sqlite3.Cursor object at 0x7f0fff6c2f10>
>>> c.execute('CREATE TABLE trReviews'\
...           '(date TEXT, OriginalText TEXT, translationSuggested TEXT)')
<sqlite3.Cursor object at 0x7f0fff6c2f10>
>>> conn.commit()
>>> conn.close()
>>> exit()

To view each contribution run the following command

>>> for row in c.execute('SELECT * FROM trReviews'):
...     print(row)

Customize to serve your model

• Your model : A checkpoint file named best.ckpt should be in the transformer older

• The source and target vocabulary named src_vocab.txt and trg_vocab.txt should be in the transformer folder.

• You don't need to change the config.yaml if you have a BPE based model.

• Change the flag if your language is not from Benin 😄 . Just replace the image in the static folder and don't forget to update the name of the file in the index.html present in the templates folder. No need to say you should change the title as well.

---

The model can be quite big and Github have restrictions on the size of your free repos. You can host your model on google drive as I do. If you want to try the English to fongbe model 😊 , you can download it here: https://drive.google.com/open?id=1PbrojtNSeZf8QpYmgJXtbcgYijVUU7Ay

I've left the vocab files in the repo. Just replace them with yours if you want to try your trained model.

Running the App

You can run the app with this commmand and start making predictions

gunicorn -b 0.0.0.0:8080 app:app

and go to localhost:8080 to interact with the app

Credits

Thanks to the Masakhane community for providing everything needed to start this project: data, code and guidance. Join us in our effort to bring Africa on the Machine Translation Map

Thanks to Julia for JoeyNMT and Slack-joey. This is based of her work. I just added a simple front end and the flask code.