deep_rl_hw2 | Trains tests a Q-network on the CartPole-v0

deep_rl_hw2 is a Python library. deep_rl_hw2 has no bugs, it has no vulnerabilities and it has low support. However deep_rl_hw2 build file is not available. You can download it from GitLab.

Trains or tests a Q-network on the CartPole-v0 or MountainCar-v0 OpenAI Gym environment. --env (string) ['CartPole-v0', 'MountainCar-v0'] name of OpenAI Gym environment to use --model_type (string) ['linear', 'dqn_1hidden', 'dqn_2hidden', 'dqn_3hidden', 'dueling_dqn'] network architecture --learning_rate (float) learning rate used during training --epsilon_start (float) beginning value of epsilon which decays over 100000 iterations to epsilon_end --epsilon_end (float) ending value of epsilon which decays from epsilon_start over 100000 iterations. --train (bool) [0, 1], default = 1. 0 if used to train a network, 1 if only testing a network --render (bool) [0, 1], default = 0. 1 to render the environment, otherwise 0 --model_path (string) path to specific tensorflow model to be used --activations (string) default = 'relu', activation used in hidden layers --replay_freq (string) [none, 'every_iter', 'every_episode'], default = 'every_iter'. We experimented with training from memory with experience replay every iteration and every episode. experience replay can be turned off when set to none.