Reduces Super Mario Maker level generation to a text generation problem.
1 - In order to generate training data, just run massconvert.sh with bash into a directory full of cdt files.
2 - feed the results as training data to a text generation machine learning model
3 - convert the result back to a level filling the necessary values by running assemble.py with a base cdt file, a generated level string file and the output file name as command line arguments.
