Trained machine shape different from input shape

Hi,

When using bob.learn.libsvm in my own project (through bob.pad.face). I sometimes got the following error:

RuntimeError: 2D `input' array should have 86 columns, matching `bob.learn.libsvm.Machine' input size, not 96

So basically the trained machine input size does not correspond to the feature size, which is quite odd ...

A first issue (bob.pad.base#19 (closed)) was open there by myself, but I guess that here is more appropriate.

Also, I tracked down the error in the C++ code, and it happens here: https://gitlab.idiap.ch/bob/bob.learn.libsvm/blob/master/bob/learn/libsvm/machine.cpp#L543

@andre.anjos any idea on this one ?

Thanks