Skip to content
A pedagogical implementation of panel apps served up on a remote machine.
Jupyter Notebook
Branch: master
Clone or download
Latest commit 18b15d4 Aug 5, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data trying to minimize memory footprint Aug 3, 2019
.gitignore ignoring .vscode Aug 2, 2019
LICENSE Initial commit Jul 24, 2019
Procfile removed --no-browser flag Jul 24, 2019
README.md add pedagogical text Aug 3, 2019
environment.yml completed modelling work Aug 3, 2019
minimal-panel.ipynb fix bug Aug 6, 2019
model-training.ipynb add pedagogical text Aug 3, 2019
requirements.txt fix holoviews version Aug 3, 2019
utils.py completed modelling work Aug 3, 2019

README.md

minimal-panel-app

A pedagogical implementation of panel apps served up on a remote machine.

See the full app here.

why this project exists

I spent a day figuring out how to make this happen at work, and decided to spend an evening consolidating my knowledge.

"how to use"

anything else interesting?

iPad development

The first version of the app was coded up entirely on an iPad, using a combination of blink Juno, and nano on my home remote server (which is nothing more than a converted gaming tower).

Web app development in Python is now doable and we can use modern tablets as a thin client!

memory usage

Deploying the HIV drug resistance model to Heroku was challenging because I had to watch out for memory and storage usage. There are 8 models to make predictions on, and loading all of them together causes memory overload on Heroku's free tier.

I got around this by pickling the models individually, and only loading them when needed. I also minimized disk usage by using gzip when pickling the files.

You can’t perform that action at this time.