Here's my music generator. The encoding script and model are in this repository. Then, I used python (Google Colab) to train the model because of the free GPU. Check out that part of the code here.
For more info, check out n8's video.
To try it, go to the google colab code and upload the model and the dataset. As a warning, you will probably need a decent understanding to tensorflow to do this. I am working on a web player currently, which will make this process a lot easier.
If you don't want to go through that agonizing process, download some of the midi files from the results folder. They aren't cherry-picked so you'll get to see what the model is really like.