Experimenting with hyper-parameter tuning in a neural-network using Jupyter and Voila
A couple of days ago, I saw a post on LinkedIn which mentioned a library called Voila that serves Jupyter (IPython) widgets in an HTML format.
I’ve been using Jupyter notebooks as part of learning data-science skills (in particular, deep-learning frameworks) for about a year now, but this was the first time I heard of widgets in the same context. This picqued my interest, because for me, Jupyter notebooks had always been Markdown + Python + Graphs/plots. Sure, you can “interact” with a Jupyter notebook by modifying your code and re-running the cells (even execute them out-of-order), but you lose that feature when you end up exporting it to static HTML. What you see in an HTML output is nothing other than the text version of your Jupyter notebook. What this means is that you can only read the code, but neither can you run nor can you modify it.