3-Dimensional Input for Multivariate Timeseries Prediction

Hi there,
Recently picked up XGBoost for a university project. Was wondering whether I am able to run multivariate time-series prediction using 3-dimensional inputs to train my model - (# of training samples, length of time, # of features). In simpler terms, I would like the model to train on multiple features/devices across a certain number of hours…is this possible?

Happy to clarify.

Many thanks.