Orange Forum • View topic - Bug in using SVM on Orange beta 0.9.5

Bug in using SVM on Orange beta 0.9.5

Report bugs (or imagined bugs).
(Archived/read-only, please use our ticketing system for reporting bugs and their discussion.)
Forum rules
Archived/read-only, please use our ticketing system for reporting bugs and their discussion.

Bug in using SVM on Orange beta 0.9.5

Postby dweaver » Fri Jul 16, 2004 21:30

I downloaded the orange-win-0.9.5.exe installer yesterday and began verifying that all my scripts from the previous orange version would work (I had been using orange v0.9.0, downloaded on 6/4/2004).

With this version, if I run the following 3 steps:

1. data = orange.ExampleTable(r"C:\Python23\Lib\site-packages\orange\doc\adult_sample.tab")
2. svm = orngSVM.BasicSVMLearner()
3. res = orngTest.crossValidation([svm], data)

I immediately get the following error:

python.exe - Application Error
The expection Illegal Instruction
An attempt was made to execute an illegal instruction.
(0xc000001d) occurred in the application at location 0x021056c0.

I'm using Windows 2000 SP3 (5.00.2195) running these steps through IPython interactive interpreter. I get a similar problem when I try to do this on OrangeCanvas.

Note also, that is is a new problem in 0.9.5 that I didn't see in 0.9.0.

Let me know if you'd like me to do more tests to pinpoint the problem.

Cheers,
Daniel

dweaver@arraybiopharma.com

Some more information on this bug

Postby Guest » Fri Jul 16, 2004 22:05

From this error message, I loaded the application into VisualStudio .NET to get more information about where the error was occuring. This revealed that the error was arising from _orngCRS.pyd. I swapped in _orngCRS.pyd from the 0.9.0 download and the SVM now works.

Mind you, this seems like an "iffy" work-around, but it's allowing me to proceed for the moment.

Any guesses about what's changed in the _orngCRS.pyd that would create this error?

Cheers,
Daniel

Postby Janez » Sat Jul 24, 2004 20:04

> Any guesses about what's changed in the _orngCRS.pyd that
> would create this error?

Yes, I guess that Aleks Jakulin, the guy behind orngCRS, was too enthusiastic when cleaning memory leaks. I'll ask him to see what's wrong when he returns from a conference (if Canadian bears don't get him first). In the meantime, I think you can safely use the older version of orngCRS. orngCRS is an independent module to some degree and is being developed separately.

Janez


Return to Bugs