Weka SMOreg and LIBSVM with linear kernel problems...

I want to test a dataset in weka using either LIBSVM with an e-SVR or SMOreg for regression. I also choose a linear kernel in both (in SMOreg i use an exponent=1 in a non normalized polykernel).

After cross-validation the root mean squared error (rmse) as well as the mean absolute error return both are equal to NaN.

Is this supposed to be an error in the values of the dataset I use (it contains no missing values)... How can I possibly handle this ? Any hints on what should I check ? The data I use for training are the appearence and shape parameter values of an Active Appearence Model created with Menpo.io using python...

Same problem occurs when I cross validate using the MLPY wrapper of Libsvm in python.