SVM Feature Selection with Multiclass Dataset from Command Line

I have a multiclass dataset and I want to apply SVM Feature Selection (SVMAttributeEval). From Weka I know that: "Attribute selection for multiclass problems is handled by ranking attributes for each class seperately using a one-vs-all method and then "dealing" from the top of each pile to give a final ranking.".
If I load the dataset using the GUI and I apply the Feature Selection than I have no problems.
Unfortunately I need to use the 10-CV (and create 10 different arff pairs of Train and Test Set). To deal with this I have already use, for binary problems, the Command Line with:

java weka.filters.supervised.attribute.AttributeSelection -E "weka.attributeSelection.SVMAttributeEval -X 1 -Y 20 -Z 0 -P 1.0E-25 -T 1.0E-10 -C 1.0 -N 0" -S "weka.attributeSelection.Ranker -T -1.7976931348623157E308 -N 500"
-b -i TrainFold0.arff
-o TrainFold0_Reduced.arff
-r TestFold0.arff
-s TestFold0_Reduced.arff

in order to "Train" the feature selection method on the original pair Train/Test and mirrors it to obtain Train_Reduced/Test_Reduced.

Now when I tried to use the same approach with my multiclass dataset I have:

SVMAttributeEval: Cannot handle multi-valued nominal attributes!

But what I don't understand is that, if I use the GUI and load for example TrainFold0.arff directly, then I can use the same method without any error or problem, What is wrong?