google-deepmind/mathematics_dataset ? reverse-engineered prompt

Build me a Python project that generates lots of school level math question and answer pairs for testing how well models can do math and reasoning.

I want it to cover a mix of problem types like algebra, arithmetic, calculus, comparison, measurement, numbers, polynomials, and probability, and to be able to print sample questions and answers from the command line. It should also support generating examples to files so I can use them for training or evaluation later.

Make the data feel varied and realistic, with easy, medium, and hard versions for training, and keep the questions and answers short like the examples in the README. If there’s a simple way to run a specific category or filter, include that too.

Please keep it easy to install and use from Python, and if you need current details from docs or the web, look them up.