Deep Q-network with replay buffer solving CartPole environment (comaptible with TensorFlow 2.0).
Although the network seems to be trained, the training result is not solid.
It need more advanced algorithms (target network, dueling, priotized buffer...) and hyper-parameter tunning.