research code
A modular system to manage research code effectively so you can move quickly while enabling reuse and collaboration.
Run code and inspect results using the same abstraction. Check out the example below ⏬
Spend more time experimenting while relying on machinable to keep things organized.
Tweak, extend, override while leveraging first-class support for Jupyter as well as the CLI.
Some research code
Running code ...
python regression.py --rate=0.1 --logs=1 --name=run-01
... and loading the corresponding results ...
python plot_regression_result.py --experiment=run-01
... are distinct and often redundant.
This means you have to manually keep track by remembering what the experiment with rate=0.1 was called.
machinable research code
Running code ...
get("regression", {"rate":0.1, "logs_": True}).launch()
... and loading the corresponding results ...
get("regression", {"rate":0.1, "logs_": True}).launch()
... are distinct but use the same abstraction.
This means no need to worry about names as machinable automatically keeps track if you ran rate=0.1 before.
machinable