How to split dataset into training and test set in MATLAB with latin hypercube sampling

I have a dataset of roughly 200 points right now and want to create a surrogate model from it in MATLAB. Herefore I apply the DTU Toolbox for the creation of the surrogate model.

My questions now is how to I prepare the dataset properly. Meaning I want to roughly split it into ~70% for the training and the other 30% for the verification of my model. My professor mentioned I could apply latin hypercube sampling in MATLAB therefore.

I found the function lhsdesign() in Matlab but as I get it it just creates a latin hypercube sample of random points with a lower and upper boundary. These points don´t represent points of my dataset tough. So is it possible to create this distribution of this specific dataset in Matlab ?

Or do I have to do a second step and try to match the random points with my dataset ?