The beautiful thing about open source is that all the detail are there for you to see!
Looking at the cdma_parameters.py file,
you can see:
pulse_training = numpy.array((1,1,1,1,-1,1,1,-1))+0j
pulse_data =numpy.array((-1,1,-1,1,-1,-1,-1,-1))+0j
so we are using 8 chips per symbol with two orthogonal codes for training and data.
You can change them and put anything you like (they better be orthogonal AND each should have good
autocorrelation properties-- or at least the training code should)
best,
Achilleas