8000 test_logistic::test_dtype_match failure on 32 bit Windows · Issue #11908 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

test_logistic::test_dtype_match failure on 32 bit Windows #11908

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
rth opened this issue Aug 24, 2018 · 9 comments
Closed

test_logistic::test_dtype_match failure on 32 bit Windows #11908

rth opened this issue Aug 24, 2018 · 9 comments
Labels
Bug help wanted

Comments

@rth
Copy link
Member
rth commented Aug 24, 2018

The test_logistic::test_dtype_match fails on 32 bit Windows as was reported in MacPython/scikit-learn-wheels#7 (comment) The implication is that for LogisticRegression, float32 and float64 input may result in somewhat different models depending on the processor architecture (and OS), which is troublesome.

This applies to Python 3.5, 3.6 but not 2.7, 3.4 and 3.7 (cf https://ci.appveyor.com/project/sklearn-wheels/scikit-learn-wheels/build/1.0.66). Linux and Mac OS are fine, and I can confirm that I can't reproduce this with a Debian unstable 32 bit Docker image.

A workaround was applied #11899 consisting in raising tolerance to 1e-2 (from 1e-6) on 32 bit Windows, but identifying the root cause of this problem would be preferable.
@rth rth mentioned this issue Aug 24, 2018
Merged
@amueller
Copy link
Member
amueller commented Sep 7, 2018

we don't care about this too much for the final release, right?

@jnothman jnothman added this to the 0.21 milestone Oct 4, 2018
@jnothman
Copy link
Member
jnothman commented Apr 9, 2019

I thought MacPython/scikit-learn-wheels#7 (comment) further suggested that repeated fits of LogisticRegression could give different solutions, at least on the Mac testing instances.

@jnothman
Copy link
Member

I suppose we'll delay this again?

@jnothman jnothman modified the milestones: 0.21, 0.22 Apr 16, 2019
@amueller
Copy link
Member

delay again?

@jnothman
Copy link
Member

Is someone available to look into these things (specifically determinism in logistic regression)?

@thomasjpfan
Copy link
Member

I will look into this.

@jnothman jnothman modified the milestones: 0.22, 0.23 Oct 31, 2019
@adrinjalali
Copy link
Member

@thomasjpfan did you get anywhere with this? Removing "blocker" since it doesn't seem to be blocking releases ;)

@thomasjpfan
Copy link
Member

I think I couldn't pin point the issue when I was looking into it. At some point I would want to get back to this when I want to venture into VM windows land.

There was "temporary fixed" in #11899 by changing the tolerance based on the bit. This is not a blocker.

@adrinjalali adrinjalali removed this from the 0.23 milestone Apr 15, 2020
@adrinjalali
Copy link
Member

We don't support <3.7 and soon <3.8, and this seems to be not reproducible, closing for now, till it appears again.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
Bug help wanted
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
5 participants
@jnothman @amueller @rth @adrinjalali @thomasjpfan
0