My second attempt at a machine-learning block-falling-game solver.
T Piece only demo: https://www.youtube.com/watch?v=zpp393LZ0XQ
Currently using a Deep Q-Learning model. Training time took about 50 hours.
Planning to upgrade hardware and rewrite the project entirely.
-- Newest progress -- https://i.imgur.com/eMNCj8E.gifv not bad I guess