arnemart / torch-rnn-web-server

torch-rnn web server

This is a simple Docker-based web server for presenting data generated using torch-rnn. It works on Dokku and probably on Heroku as well.

It allows you to do stupid shit like this page that creates new book titles for Harlequin Romance.

Out of the box this runs a simple node.js-powered web server that fetches data from torch-rnn, caches it and then presents it in a web page. There is also a script for tweeting output from the neural network.

How to create checkpoint files

Once you have your input data, refer the documentation for torch-rnn and/or docker-torch-rnn where all this is thoroughly explained.

How to use

The app ships with a sample checkpoint from torch-rnn ()trained on all book titles from Harlequin Romance), which is referenced in the index.js file by default. Replace this with your own checkpoint file when you have one.

To build and run the docker image:

1. cd torch-rnn-web-server
2. Run docker build . -- the output of this command should end with "Successfully built IMAGE_ID>"
3. Copy the IMAGE_ID and run docker run -p 8000:80 IMAGE_ID
4. You can now point your browser to http://localhost:8000 and view the neural network output in all its glory.
5. (To stop the docker container, run docker ps to get the container_id and then run docker stop CONTAINER_ID)

The data processing that happens in index.js and all the HTML/CSS in template.html can of course be edited to your heart's content.

Why to use

That's not for me to say