Improve computational efficiency or speed up training process with GPU is applicable

[huanyi89]

Hello team,

Thank you for the great package.
I have some struggle for using this package because of my dataset. The main issue is the speed.
I have data set with at least 50+ features (dimension), and 10K+ observations (rows), which causes a long time for structure learning and Bayesian Network.

I wonder if there is any opportunity to improve the computational efficiency, and
I wonder since this package uses Pytorch, is that possible to leverage GPU for learning?

If yes, is there any guidance for me to follow?

Thank you

[oentaryorj]

This has been implemented in this commit and will be made available in the next release.