def test_multi_agent_dqn():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, algorithm='DQN', very_fast=False, visualize = True, max_iterate=500,
file_or_folder_to_load='../modeldata/deep_q_3G_3Z_5000_times_2019_01_27_08_00.h5',
mode='evaluate', epsilon_decrease='LINEAR', map_name='3G_3Z', layers=[100,100])
trainer.train()
def train_multi_agent2():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, algorithm='DeepSarsa', very_fast=True, visualize = False, max_iterate=3000,
mode='train',epsilon_decrease='INVERSE_SQRT',map_name='3G_4Z',layers=[100,100],
export_per=1000,last_action_state_also_state=False)
trainer.train()
def train_multi_agent():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, epsilon_decay=InvSqrtDecay(1.0), algorithm='DeepSarsa', very_fast=True, visualize = False, max_iterate=5000,
#file_to_load='../modeldata/DeepSarsa3G5Z/deep_sarsa_3G_5Z_5000_times_2019_01_31_08_54.h5',
mode='train',map_name='3G_4Z',layers=[100,100],
export_per=1000,last_action_state_also_state=True)
trainer.train()
def train_multi_agent_dqn():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, algorithm='DQN', very_fast=True, visualize = False, max_iterate=5000,
#file_to_load='../modeldata/deep_q_3G_3Z_500_times_2019_01_26_23_12.h5',
mode='train',epsilon_decrease='LINEAR',map_name='3G_4Z',layers=[100,100],
export_per=250)
trainer.train()
def test_multi_agent():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, algorithm='DeepSarsa', very_fast=False, visualize = True, max_iterate=100,
file_or_folder_to_load='../modeldata/deep_sarsa_3G_6Z_2019_02_06_00_48/deep_sarsa_3G_6Z_4000_times_2019_02_06_11_31.h5',
mode='evaluate', epsilon_decrease='LINEAR', map_name='3G_6Z', layers=[100,100]
, last_action_state_also_state = False)
trainer.train()
def train_multi_agent_a2c_eligibility():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, epsilon_decay=ConstantEpsilon(0.0), algorithm='A2C', very_fast=True, visualize = False,
max_iterate=50000,
mode='train',map_name='3G_6Z', actor_layers=[100, 100], critic_layers=[100, 100],
export_per=1000,last_action_state_also_state=False, eligibility_trace=True)
trainer.train(do_test_during_train=False)
def train_multi_agent_with_eligibility():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, epsilon_decay=InvSqrtDecay(1.0), algorithm='DeepSarsa', very_fast=True, visualize = False, max_iterate=5000,
mode='train',map_name='3G_6Z',layers=[100, 100],
export_per=500,last_action_state_also_state=False,
eligibility_trace=True)
trainer.train(do_test_during_train=False)
def test_multi_agent_multiple_models():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, epsilon_decay=ConstantEpsilon(0.0), algorithm='DeepSarsa', very_fast=True, visualize=False, max_iterate=100,
file_or_folder_to_load='../modeldata/multi_test',
mode='evaluate_multiple_models', map_name='3G_6Z', layers=[100, 100]
, last_action_state_also_state=False)
trainer.evaluate_multiple(test_file_path="deep_sarsa_3G_6Z_2019_02_06_23_53", test_zero=True, test_iter=100)
def test_multi_agent():
s = socketClient()
s.accessToServer()
trainer = MultiAgentTrainer(s, epsilon_decay=ConstantEpsilon(0.0), algorithm='DeepSarsa', very_fast=True, visualize = False, max_iterate=100,
file_or_folder_to_load='../modeldata/deep_sarsa_3G_6Z_2019_02_06_23_53/deep_sarsa_3G_6Z_4000_times_2019_02_07_08_54.h5',
mode='evaluate', map_name='3G_6Z', layers=[100,100]
, last_action_state_also_state = False)
#trainer.evaluate(100)
trainer.train()