MEMORY_CAPACITY = int(1e+5)
# simulator steps for learning interval
LEARN_FREQ = 4
'''Environment Settings'''
# number of environments for C51
N_ENVS = args.nenv
# Total simulation step
STEP_NUM = int((2e+7)+2)
# gamma for MDP
GAMMA = 0.99
# visualize for agent playing
RENDERING = False
# openai gym env name
ENV_NAME = args.games+'NoFrameskip-v4'
env = SubprocVecEnv([wrap_cover(ENV_NAME,args.seed+i) for i in range(N_ENVS)])
N_ACTIONS = env.action_space.n
N_STATES = env.observation_space.shape
'''Training settings'''
# check GPU usage
USE_GPU = torch.cuda.is_available()
print('USE GPU: '+str(USE_GPU))
# mini-batch size
BATCH_SIZE = 64
# learning rage
LR = args.lr
# epsilon-greedy
EPSILON = 1.0
'''Save&Load Settings'''
LEARN_FREQ = 4
# quantile and option numbers for QUOTA
N_QUANT = 200
N_OPTIONS = 10
'''Environment Settings'''
# number of environments for C51
N_ENVS = 16
# Total simulation step
STEP_NUM = int(1e+8)
# gamma for MDP
GAMMA = 0.99
# visualize for agent playing
RENDERING = False
# openai gym env name
ENV_NAME = args.games + 'NoFrameskip-v4'
env = SubprocVecEnv([wrap_cover(ENV_NAME) for i in range(N_ENVS)])
N_ACTIONS = env.action_space.n
N_STATES = env.observation_space.shape
'''Training settings'''
# check GPU usage
USE_GPU = torch.cuda.is_available()
print('USE GPU: ' + str(USE_GPU))
# mini-batch size
BATCH_SIZE = 32
# learning rage
LR = 1e-4
# epsilon-greedy
EPSILON = 1.0
EPSILON_O = 1.0
# option paramater
Target_beta = 0.01
MEMORY_CAPACITY = int(1e+5)
# simulator steps for learning interval
LEARN_FREQ = 4
'''Environment Settings'''
# number of environments for C51
N_ENVS = 16
# Total simulation step
STEP_NUM = int((1e+7) + 2)
# gamma for MDP
GAMMA = 0.99
# visualize for agent playing
RENDERING = False
# openai gym env name
ENV_NAME = args.games + 'NoFrameskip-v4'
env = SubprocVecEnv(
[wrap_cover(ENV_NAME, args.seed + i) for i in range(N_ENVS)])
N_ACTIONS = env.action_space.n
N_STATES = env.observation_space.shape
'''Training settings'''
# check GPU usage
USE_GPU = torch.cuda.is_available()
print('USE GPU: ' + str(USE_GPU))
# mini-batch size
BATCH_SIZE = 64
# learning rage
LR = 1e-4
# epsilon-greedy
EPSILON = 1.0
'''Save&Load Settings'''
# check save/load
SAVE = True