A3C

Example

 from rlzoo.common.env_wrappers import build_env
 from rlzoo.common.utils import call_default_params
 from rlzoo.algorithms import A3C

 AlgName = 'A3C'
 EnvName = 'PongNoFrameskip-v4'
 EnvType = 'atari'

 # EnvName = 'Pendulum-v0'  # only continuous action
 # EnvType = 'classic_control'

 # EnvName = 'BipedalWalker-v2'
 # EnvType = 'box2d'

 # EnvName = 'Ant-v2'
 # EnvType = 'mujoco'

 # EnvName = 'FetchPush-v1'
 # EnvType = 'robotics'

 # EnvName = 'FishSwim-v0'
 # EnvType = 'dm_control'

 number_workers = 2  # need to specify number of parallel workers
 env = build_env(EnvName, EnvType, nenv=number_workers)
 alg_params, learn_params = call_default_params(env, EnvType, AlgName)
 alg = eval(AlgName+'(**alg_params)')
 alg.learn(env=env,  mode='train', render=False, **learn_params)
 alg.learn(env=env,  mode='test', render=True, **learn_params)

Asychronous Advantage Actor-Critic

class rlzoo.algorithms.a3c.a3c.A3C(net_list, optimizers_list, entropy_beta=0.005)

learn(env, train_episodes=1000, test_episodes=10, max_steps=150, render=False, n_workers=1, update_itr=10, gamma=0.99, save_interval=500, mode='train', plot_func=None)

Parameters:

• env – a list of same learning environments
• train_episodes – total number of episodes for training
• test_episodes – total number of episodes for testing
• max_steps – maximum number of steps for one episode
• render – render or not
• n_workers – manually set number of workers
• update_itr – update global policy after several episodes
• gamma – reward discount factor
• save_interval – timesteps for saving the weights and plotting the results
• mode – train or test
• plot_func – additional function for interactive module

Default Hyper-parameters

rlzoo.algorithms.a3c.default.atari(env, default_seed=True)

rlzoo.algorithms.a3c.default.box2d(env, default_seed=True)

rlzoo.algorithms.a3c.default.classic_control(env, default_seed=True)

rlzoo.algorithms.a3c.default.dm_control(env, default_seed=True)

rlzoo.algorithms.a3c.default.mujoco(env, default_seed=True)

rlzoo.algorithms.a3c.default.rlbench(env, default_seed=True)

rlzoo.algorithms.a3c.default.robotics(env, default_seed=True)