mummify is a version control tool for machine learning. It's simple, fast, and designed for model prototyping.

Add mummify.log(<string>) to the bottom of a machine learning model:

from sklearn.datasets import load_wine
from sklearn.neighbors import KNeighborsClassifier

import mummify

data = load_wine()
X, y = data.data, data.target

model = KNeighborsClassifier(n_neighbors=4)
model.fit(X, y)
accuracy = round(model.score(X, y), 4)

mummify.log(f'Accuracy: {accuracy}')

Run the model at the command line:

python model.py

Edit the model to implement another algorithm:

...
model = LogisticRegression()
model.fit(X, y)
accuracy = round(model.score(X, y), 4)

mummify.log(f'Accuracy: {accuracy}')

Inspect model history at the command line with:

mummify history

And peek at the logged messages at the command line with:

cat mummify.log

Switch to an earlier version of the model:

mummify switch <id>

mummify will persist snapshots and the mummify.log file between switches.

pip install mummify

For feature requests or bug reports, please use Github Issues