HN Explorer

Ask HN: How do you keep track of your experiment scripts?

2 points by Scea91 ↗ HN
When doing research I often write a single script and then run it multiple times, each time tweaking some part of the script or adjusting constants. I need to keep track of all of these versions to be able to reproduce the results.

Currently I am making sure that before each experiment, the current state of the code is commited to git and I tag it with experiment's name. If I need to expand on past experiment I branch out from a commit in history.

I would appreciate some insights from other people on how they handle this.

1 comment

[ 2.9 ms ] story [ 11.3 ms ] thread
visarga
I was thinking about the same problem, would appreciate input.

I am unhappy to run experiments without forming some kind of map to help me navigate the various choices. Most runs end up forgotten, and that's very inefficient.