Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-11 Thread Alexandre Gramfort
chris please open an issue. it also seems like the issue appears with lsqr but not with other solvers. Alex -- Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer Achieve PCI DSS 3.0 Compliant Status with Out

Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-11 Thread Chris Holdgraf
Ahhh, that is a good point Joel, I hadn't thought about that. I think you may be right, check out the updated gist . I also wrote in some objgraph code to print the backrefs before and after the "fit" method is called using Ridge. He

Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-11 Thread Joel Nothman
Is there reason to believe this is associated with grid search, and not just a memory leak in Ridge? On 11 October 2014 20:42, Alexandre Gramfort < [email protected]> wrote: > Indeed I can reproduce the problem. > > Maybe @MechCoder can have a look. > > out of curiosity why

Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-11 Thread Alexandre Gramfort
Indeed I can reproduce the problem. Maybe @MechCoder can have a look. out of curiosity why don't you use RidgeCV? Alex -- Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer Achieve PCI DSS 3.0 Compliant Sta

Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-10 Thread Chris Holdgraf
I recreated it with some fake data, here's a short gist to show what I'm talking about: http://nbviewer.ipython.org/gist/choldgraf/6a7be7866f2a3a3d3f98 On Fri, Oct 10, 2014 at 1:29 PM, Chris Holdgraf wrote: > Yes - this is with the latest scikit-learn. Also, I'm using n_jobs==1, so > there shou

Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-10 Thread Chris Holdgraf
Yes - this is with the latest scikit-learn. Also, I'm using n_jobs==1, so there shouldn't be any memmapping anyway, right? Which version of scikit-learn? Have you tried with 0.15.2? Your data should automatically get memory mapped to share some input data with the 'n_jobs' worker processes. --

Re: [Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-10 Thread Olivier Grisel
Which version of scikit-learn? Have you tried with 0.15.2? Your data should automatically get memory mapped to share some input data with the 'n_jobs' worker processes. -- Olivier -- Meet PCI DSS 3.0 Compliance Requirem

[Scikit-learn-general] GridSearchCV using a ton of memory

2014-10-10 Thread Chris Holdgraf
Hey all - I'm running into some memory issues with GridSearchCV and I wonder if anyone can give an intuition as to why. I'm cross-validating alpha parameters for Ridge regression. I'm trying 8 different parameters. My inputs are 2400x1900 (~370MB) in size. When I run %memit model.fit(X, y) alo