This is DQN implemented with tensorflow 2. DQN implemented in two version. The original DQN (Nature 2015) and multi-step DQN. The experiment has done under the OpenAI gym cartpole-v1 environment.

• Python 3.8.2
• tensorflow 2.2.0rc2
• matplotlib 3.2.1

Training only original DQN.

> python script.py orgDQN

Training original DQN and multi-step DQN in a sequence.

> python script.py orgDQN multistep

If the average step count exceed 475, the training ends early.