def __init__(self):
self.gamma = 0.99
self.critic_learning_rate = 0.0002
self.actor_learning_rate = 0.0001
self.tau = 0.001
self.batch_size = 64
self.update_frequency = 4
self.exploration = RLAgents.DecayConst(0.1, 0.1)
self.experience_replay_size = 200000
def __init__(self):
self.gamma = 0.99
self.critic_learning_rate = 0.0002
self.actor_learning_rate = 0.0001
self.forward_learning_rate = 0.0002
self.autoencoder_learning_rate = 0.0002
self.tau = 0.001
self.beta1 = 0.1
self.beta2 = 0.1
self.episodic_memory_size = 128
self.batch_size = 64
self.update_frequency = 4
self.exploration = RLAgents.DecayConst(0.1, 0.1)
self.experience_replay_size = 200000
import RLAgents
import matplotlib.pyplot as plt
result_path = "./results/"
files = []
files.append("./models/ddpg_baseline/run_0/result/result.log")
files.append("./models/ddpg_baseline/run_1/result/result.log")
files.append("./models/ddpg_baseline/run_2/result/result.log")
files.append("./models/ddpg_baseline/run_3/result/result.log")
files.append("./models/ddpg_baseline/run_4/result/result.log")
files.append("./models/ddpg_baseline/run_5/result/result.log")
files.append("./models/ddpg_baseline/run_6/result/result.log")
files.append("./models/ddpg_baseline/run_7/result/result.log")
ddpg_baseline = RLAgents.RLStatsCompute(files)
files = []
files.append("./models/ddpg_curiosity/run_0/result/result.log")
files.append("./models/ddpg_curiosity/run_1/result/result.log")
files.append("./models/ddpg_curiosity/run_2/result/result.log")
files.append("./models/ddpg_curiosity/run_3/result/result.log")
files.append("./models/ddpg_curiosity/run_4/result/result.log")
files.append("./models/ddpg_curiosity/run_5/result/result.log")
files.append("./models/ddpg_curiosity/run_6/result/result.log")
files.append("./models/ddpg_curiosity/run_7/result/result.log")
ddpg_curiosity_rnd = RLAgents.RLStatsCompute(files)
'''
files = []
files.append("./models/ddpg_entropy/run_0/result/result.log")
files.append("./models/ddpg_entropy/run_1/result/result.log")
import gym
import pybullet_envs
import numpy
import time
import RLAgents
import models.ddpg_baseline.model.src.model_critic as ModelCritic
import models.ddpg_baseline.model.src.model_actor as ModelActor
import models.ddpg_baseline.model.src.config as Config
path = "models/ddpg_baseline/model/"
env = gym.make("AntBulletEnv-v0")
agent = RLAgents.AgentDDPG(env, ModelCritic, ModelActor, Config)
max_iterations = 4 * (10**6)
trainig = RLAgents.TrainingIterations(env, agent, max_iterations, path, 1000)
trainig.run()
'''
agent.load(path)
agent.disable_training()
while True:
reward, done = agent.main()
time.sleep(0.01)
'''
import gym
import gym_aeris
import numpy
import time
import RLAgents
import models.ddpg_curiosity.model.src.model_critic as ModelCritic
import models.ddpg_curiosity.model.src.model_actor as ModelActor
import models.ddpg_curiosity.model.src.model_forward as ModelForward
import models.ddpg_curiosity.model.src.model_forward_target as ModelForwardTarget
import models.ddpg_curiosity.model.src.config as Config
path = "models/ddpg_curiosity/model/"
env = gym.make("TargetNavigate-v0", render = False)
agent = RLAgents.AgentDDPGCuriosity(env, ModelCritic, ModelActor, ModelForward, ModelForwardTarget, Config)
max_iterations = 1*(10**6)
trainig = RLAgents.TrainingIterations(env, agent, max_iterations, path, 10000)
trainig.run()
'''
agent.load(path)
agent.disable_training()
while True:
reward, done = agent.main()
time.sleep(0.01)
'''
import gym
import gym_line_follower
import numpy
import time
import RLAgents
import models.ppo_baseline_continuous.src.model as Model
import models.ppo_baseline_continuous.src.config as Config
path = "models/ppo_baseline_continuous/"
config = Config.Config()
envs = RLAgents.MultiEnvSeq('LineFollower-v0', None, config.actors)
agent = RLAgents.AgentPPOContinuous(envs, Model, Config)
max_iterations = 1 * (10**6)
trainig = RLAgents.TrainingIterations(envs, agent, max_iterations, path, 1000)
trainig.run()
'''
agent.load(path)
agent.disable_training()
while True:
agent.main()
envs.render(0)
'''
import gym
import pybullet_envs
import numpy
import time
import RLAgents
import models.ppo_baseline.model.src.model as Model
import models.ppo_baseline.model.src.config as Config
path = "models/ppo_baseline/model/"
config = Config.Config()
envs = RLAgents.MultiEnvSeq("AntBulletEnv-v0", None, config.actors)
#envs.render(0)
agent = RLAgents.AgentPPOContinuous(envs, Model, Config)
max_iterations = 1 * (10**6)
trainig = RLAgents.TrainingIterations(envs, agent, max_iterations, path, 1000)
trainig.run()
'''
agent.load(path)
agent.disable_training()
while True:
reward, done = agent.main()
time.sleep(0.01)
'''
import gym
import gym_aeris
import numpy
import time
import RLAgents
import models.ddpg_baseline.model.src.model_critic as ModelCritic
import models.ddpg_baseline.model.src.model_actor as ModelActor
import models.ddpg_baseline.model.src.config as Config
path = "models/ddpg_baseline/model/"
env = gym.make("FoodGatheringAdvanced-v0", render=True)
agent = RLAgents.AgentDDPG(env, ModelCritic, ModelActor, Config)
max_iterations = 4 * (10**6)
#trainig = RLAgents.TrainingIterations(env, agent, max_iterations, path, 1000)
#trainig.run()
agent.load(path)
agent.disable_training()
while True:
reward, done = agent.main()
time.sleep(0.01)
duration=duration,
loop=0)
if __name__ == '__main__':
"""
"""
start = time.time()
EpsImagePlugin.gs_windows_binary = r'C:\Program Files\gs\gs9.54.0\bin\gswin64c'
##np.seterr(all='raise')
random.seed('sdsc8006')
layouts = ['small', 'medium']
pacmans = {
'MC': RLAgents.MonteCarloAgent(eps0=1e1, gamma=1),
'TD': TDAgent(eps0=1e1, gamma=1),
'QL': RLAgents.QLearningAgent(eps0=1, gamma=1, alpha=1e-4),
# alpha for w update, beta for theta update
'AC': RLAgents.ActorCriticAgent(gamma=1, alpha=1e-4, beta=1e-4),
}
argsList, options = readCommand(sys.argv[1:])
if not options.noRun:
for args in argsList:
print(args)
runGames(**args)
if options.testAll:
plotAll(layouts,
import gym
import gym_aeris
import numpy
import time
import RLAgents
import models.ddpg_entropy.model.src.model_critic as ModelCritic
import models.ddpg_entropy.model.src.model_actor as ModelActor
import models.ddpg_entropy.model.src.model_forward as ModelForward
import models.ddpg_entropy.model.src.model_forward_target as ModelForwardTarget
import models.ddpg_entropy.model.src.model_ae as ModelAutoencoder
import models.ddpg_entropy.model.src.config as Config
path = "models/ddpg_entropy/model/"
env = gym.make("TargetNavigate-v0", render=False)
agent = RLAgents.AgentDDPGEntropy(env, ModelCritic, ModelActor, ModelForward,
ModelForwardTarget, ModelAutoencoder, Config)
max_iterations = 1 * (10**6)
trainig = RLAgents.TrainingIterations(env, agent, max_iterations, path, 10000)
trainig.run()
'''
agent.load(path)
agent.disable_training()
while True:
reward, done = agent.main()
time.sleep(0.01)
'''