Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

sklearn shouldn't warn when unpickling an Estimator with a different patch version

See original GitHub issue

Description

I pickled an estimator with sklearn 0.22.0 and I’m trying to unpickle it with 0.22.1.

Expected Results

No warning since this is a patch version change, there should be no breaking changes.

Actual Results

I got the following warning:

UserWarning: Trying to unpickle estimator MinMaxScaler from version 0.22 when using version 0.22.1. This might lead to breaking code or invalid results. Use at your own risk.

Issue Analytics

State:
Created 4 years ago
Comments:8 (4 by maintainers)

Top GitHub Comments

1reaction

victornoelcommented, Jan 9, 2020

@amueller thanks for the clarification.

Basically your point is that pickling is not a public API. The reason I created this issue is that because from my usage, I believed that pickling of models was part of the public API. It seems I was mistaken 😃

Actually you do have documentation on this here and after reading it (I didn’t before), I now believe there exist better alternative than pickling to achieve model persistence in case backward compatibility is important.

0reactions

tinaroootcommented, Mar 16, 2022

Fuck, just changing a version at random will kill people when migrating code