def runModel():
#Parameters for the game
agent1 = AgentA2C.AgentA2C([False, 1.0]) #training agent
agent2 = AgentA2C.AgentA2C([False, 1.0])
agent3 = AgentA2C.AgentA2C([False, 1.0])
agent4 = AgentA2C.AgentA2C([False, 1.0])
# if training specific agents
playersAgents = [agent1, agent2, agent3, agent4]
reward = RewardOnlyWinning.RewardOnlyWinning()
numRuns = 10 #Amount of runs
numGames = 100 # amount of games per run
experimentDescriptor = "Testing_NewPlot"
A2cActor_1 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0Choice_ Scratch - Second Best - Best - _2020-03-26_19:11:39.863435/Model/actor_iteration_999_Player_0.hd5"
A2cCritic_1 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0Choice_ Scratch - Second Best - Best - _2020-03-26_19:11:39.863435/Model/critic_iteration_999_Player_0.hd5"
A2cActor_4 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_4_Training_Best_Agent_1Choice_ Best - Scratch - Scratch - _2020-03-26_19:27:58.464895/Model/actor_iteration_999_Player_0.hd5"
A2cCritic_4 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_4_Training_Best_Agent_1Choice_ Best - Scratch - Scratch - _2020-03-26_19:27:58.464895/Model/critic_iteration_999_Player_0.hd5"
A2cActor_9 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_1_Training_Best_Agent_2_2020-03-26_17:33:51.517296/Model/actor_iteration_999_Player_1.hd5"
A2cCritic_9 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_2_Training_Best_Agent_1_2020-03-26_17:42:24.306637/Model/actor_iteration_999_Player_1.hd5"
A2cActor_r = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Random/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-27_14:14:16.796731/Model/actor_iteration_999_Player_0.hd5"
A2cCritic_r = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Random/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-27_14:14:16.796731/Model/critic_iteration_999_Player_0.hd5"
loadModelAgent1 = [
A2cActor_1, A2cCritic_1
] #""#""#"" #DQLModel #[actorModelA2C,criticModelA2c] #[actorModelDDPG,criticModelDDPG]
loadModelAgent2 = [A2cActor_4,
A2cCritic_4] #""#""# ""#[actorModel,criticModel]
loadModelAgent3 = [A2cActor_9, A2cCritic_9] #""#""# ""
loadModelAgent4 = [A2cActor_r, A2cCritic_r] #""#DQLModelr#""#""# ""
#
loadModel = [
loadModelAgent4, loadModelAgent3, loadModelAgent1, loadModelAgent2
] #indicate where the saved model is
# loadModel = [loadModelAgent3, loadModelAgent2, loadModelAgent1,
# loadModelAgent4] # indicate where the saved model is
#
# loadModel = "" #indicate where the saved model is
# #Parameters for controling the experiment
isLogging = False # Logg the experiment
isPlotting = True # plot the experiment
plotFrequency = 1000 # plot the plots every X games
createDataset = True # weather to save the dataset
saveExperimentsIn = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/" # Directory where the experiment will be saved
winsP1 = []
winsP2 = []
winsP3 = []
winsP4 = []
for i in range(numRuns):
metrics = ChefsHatExperimentHandler.runExperiment(
numGames=numGames,
playersAgents=playersAgents,
experimentDescriptor=experimentDescriptor,
isLogging=isLogging,
isPlotting=isPlotting,
plotFrequency=plotFrequency,
createDataset=createDataset,
saveExperimentsIn=saveExperimentsIn,
loadModel=loadModel,
rewardFunction=reward)
# Player1 - agent
p1 = metrics[2]
p2 = metrics[3]
p3 = metrics[4]
p4 = metrics[5]
winsP1.append(p1[0])
winsP2.append(p2[0])
winsP3.append(p3[0])
winsP4.append(p4[0])
print("Metrics:" + str(metrics))
plotVictoriesTotal(winsP1, winsP2, winsP3, winsP4, numGames,
experimentDescriptor, saveExperimentsIn)
def runModel():
#Parameters for the game
agent1 = AgentDQL.AgentDQL([False, 1.0]) #training agent
agent2 = AgentA2C.AgentA2C([False, 1.0])
agent3 = AgentPPO.AgentPPO([False, 1.0])
agent4 = AgentRandom.AgentRandom(AgentRandom.DUMMY_RANDOM)
# if training specific agents
playersAgents = [agent1, agent2, agent3, agent4]
reward = RewardOnlyWinning.RewardOnlyWinning()
numRuns = 1 #Amount of runs
numGames = 100 # amount of games per run
experimentDescriptor = "Testing_NoTraining_newPlot"
DQLModel = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0_2020-03-26_17:26:18.198600/Model/actor_iteration_999_Player_2.hd5"
A2cActor = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0Choice_ Scratch - Second Best - Best - _2020-03-26_19:11:39.863435/Model/actor_iteration_999_Player_0.hd5"
A2cCritic = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0Choice_ Scratch - Second Best - Best - _2020-03-26_19:11:39.863435/Model/critic_iteration_999_Player_0.hd5"
PPOActor = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/PPO/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_9_Training_Best_Agent_3Choice_ Scratch - Scratch - Scratch - _2020-03-26_22:27:02.430568/Model/actor_iteration_999_Player_3.hd5"
PPOCritic = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/PPO/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'A2C', 'A2C', 'A2C']_Reward_OnlyWinning_Training_GameExperimentNumber_9_Training_Best_Agent_3Choice_ Scratch - Scratch - Scratch - _2020-03-26_22:27:02.430568/Model/critic_iteration_999_Player_3.hd5"
loadModelAgent1 = DQLModel #""#""#"" #DQLModel #[actorModelA2C,criticModelA2c] #[actorModelDDPG,criticModelDDPG]
loadModelAgent2 = [A2cActor,
A2cCritic] #""#""# ""#[actorModel,criticModel]
loadModelAgent3 = [PPOActor, PPOCritic] #""#""# ""
loadModelAgent4 = "" #""#DQLModelr#""#""# ""
#
# loadModel = [loadModelAgent1,loadModelAgent2, loadModelAgent3, loadModelAgent4] #indicate where the saved model is
loadModel = [
loadModelAgent1, loadModelAgent2, loadModelAgent3, loadModelAgent4
] # indicate where the saved model is
#
# loadModel = "" #indicate where the saved model is
# #Parameters for controling the experiment
isLogging = False # Logg the experiment
isPlotting = False # plot the experiment
plotFrequency = 1 # plot the plots every X games
createDataset = True # weather to save the dataset
saveExperimentsIn = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/All/AllQValues_beforetraining/" # Directory where the experiment will be saved
winsP1 = []
winsP2 = []
winsP3 = []
winsP4 = []
qvalues = []
for a in range(len(playersAgents)):
qvalues.append([])
for i in range(numRuns):
metrics = ChefsHatExperimentHandler.runExperiment(
numGames=numGames,
playersAgents=playersAgents,
experimentDescriptor=experimentDescriptor,
isLogging=isLogging,
isPlotting=isPlotting,
plotFrequency=plotFrequency,
createDataset=createDataset,
saveExperimentsIn=saveExperimentsIn,
loadModel=loadModel,
rewardFunction=reward)
# Player1 - agent
p1 = metrics[2]
p2 = metrics[3]
p3 = metrics[4]
p4 = metrics[5]
winsP1.append(p1[0])
winsP2.append(p2[0])
winsP3.append(p3[0])
winsP4.append(p4[0])
for a in range(len(playersAgents) - 1):
qvalues[a].append(metrics[a + 2][-1])
# print ("Metrics:" + str(metrics))
plotVictoriesTotal(winsP1, winsP2, winsP3, winsP4, numGames,
experimentDescriptor, saveExperimentsIn)
def runModel():
#Parameters for the game
agent1 = AgentA2C.AgentA2C([False, 1.0]) #training agent
agent2 = AgentRandom.AgentRandom(AgentRandom.DUMMY_RANDOM)
agent3 = AgentRandom.AgentRandom(AgentRandom.DUMMY_RANDOM)
agent4 = AgentRandom.AgentRandom(AgentRandom.DUMMY_RANDOM)
# if training specific agents
playersAgents = [agent1, agent2, agent3, agent4]
reward = RewardOnlyWinning.RewardOnlyWinning()
numRuns = 1 #Amount of runs
numGames = 100 # amount of games per run
experimentDescriptor = "Testing_NewPlot_Stack"
A2CActor = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Random/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-27_14:14:16.796731/Model/actor_iteration_999_Player_0.hd5"
A2CCritic = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Random/1000/Player_4_Cards_11_games_1000TrainingAgents_['A2C', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-27_14:14:16.796731/Model/critic_iteration_999_Player_0.hd5"
loadModelAgent1 = [
A2CActor, A2CCritic
] #""#[actorModelDDPG,criticModelDDPG]#DQLModel#""# #"" #""#""#DQLModel#""#[actorModelDDPG,criticModelDDPG] ##""#[actorModelA2C,criticModelA2C] #[actorModelA2C,criticModelA2C] #DQLModel #[actorModelA2C,criticModelA2c] #[actorModelDDPG,criticModelDDPG]
loadModelAgent2 = "" #DQLModel #"" #DQLModel
loadModelAgent3 = "" #[actorModelDDPG,criticModelDDPG]
loadModelAgent4 = ""
#
# loadModel = [loadModelAgent1,loadModelAgent2, loadModelAgent3, loadModelAgent4] #indicate where the saved model is
loadModel = [
loadModelAgent1, loadModelAgent2, loadModelAgent3, loadModelAgent4
] #indicate where the saved model is
#
# loadModel = "" #indicate where the saved model is
# #Parameters for controling the experiment
isLogging = False #Logg the experiment
isPlotting = False #plot the experiment
plotFrequency = 1 #plot the plots every X games
createDataset = True # weather to save the dataset
saveExperimentsIn = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/A2C/Random/1000/AllQValues" # Directory where the experiment will be saved
winsP1 = []
winsP2 = []
winsP3 = []
winsP4 = []
for i in range(numRuns):
metrics = ChefsHatExperimentHandler.runExperiment(
numGames=numGames,
playersAgents=playersAgents,
experimentDescriptor=experimentDescriptor,
isLogging=isLogging,
isPlotting=isPlotting,
plotFrequency=plotFrequency,
createDataset=createDataset,
saveExperimentsIn=saveExperimentsIn,
loadModel=loadModel,
rewardFunction=reward)
# Player1 - agent
p1 = metrics[2]
p2 = metrics[3]
p3 = metrics[4]
p4 = metrics[5]
winsP1.append(p1[0])
winsP2.append(p2[0])
winsP3.append(p3[0])
winsP4.append(p4[0])
# print ("Metrics:" + str(metrics))
plotVictoriesTotal(winsP1, winsP2, winsP3, winsP4, numGames,
experimentDescriptor, saveExperimentsIn)
def runModel():
#Parameters for the game
agent1 = AgentDQL.AgentDQL([False, 1.0]) #training agent
agent2 = AgentDQL.AgentDQL([False, 1.0])
agent3 = AgentDQL.AgentDQL([False, 1.0])
agent4 = AgentDQL.AgentDQL([False, 1.0])
# if training specific agents
playersAgents = [agent1, agent2, agent3, agent4]
reward = RewardOnlyWinning.RewardOnlyWinning()
numRuns = 10 #Amount of runs
numGames = 100# amount of games per run
experimentDescriptor = "Testing_50x1000"
# DQLModel1 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_1_Training_Best_Agent_2_2020-03-26_17:33:51.517296/Model/actor_iteration_999_Player_1.hd5"
# DQLModel2 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_2_Training_Best_Agent_1_2020-03-26_17:42:24.306637/Model/actor_iteration_999_Player_1.hd5"
# DQLModel3 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/Old/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_3_Training_Best_Agent_2_2020-03-26_00:13:55.479460/Model/actor_iteration_999_Player_1.hd5"
# DQLModelr = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Random/Player_4_Cards_11_games_3000TrainingAgents_['DQL', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-25_18:15:32.076987/Model/actor_iteration_2999_Player_0.hd5"
#
#
#
# DQLModel0 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0_2020-03-26_17:26:18.198600/Model/actor_iteration_999_Player_2.hd5"
# DQLModel4 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_4_Training_Best_Agent_3_2020-03-26_18:03:32.220659/Model/actor_iteration_999_Player_2.hd5"
# DQLModel9 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/Player_4_Cards_11_games_3000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0Choice_ Scratch - Second Best Agents - Second Best Agents - _2020-03-26_20:28:12.574082/Model/actor_iteration_2999_Player_0.hd5"
# DQLModelr = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Random/Player_4_Cards_11_games_3000TrainingAgents_['DQL', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-25_18:15:32.076987/Model/actor_iteration_2999_Player_0.hd5"
DQLModel0 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000x50/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_0_Training_Best_Agent_0Choice_SecondBestAgents-SecondBestAgents-Scratch-_2020-03-27_01:55:16.105188/Model/actor_iteration_999_Player_2.hd5"
DQLModel4 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000x50/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_15_Training_Best_Agent_2Choice_SecondBestAgents-BestAgents-BestAgents-_2020-03-27_04:50:05.205901/Model/actor_iteration_999_Player_2.hd5"
DQLModel9 = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000x50/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DQL', 'DQL', 'DQL']_Reward_OnlyWinning_Training_GameExperimentNumber_30_Training_Best_Agent_0Choice_BestAgents-BestAgents-BestAgents-_2020-03-27_12:18:40.723555/Model/actor_iteration_999_Player_2.hd5"
DQLModelr = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Random/1000/Player_4_Cards_11_games_1000TrainingAgents_['DQL', 'DUMMY_RANDOM', 'DUMMY_RANDOM', 'DUMMY_RANDOM']_Reward_OnlyWinning_Training_2020-03-27_14:26:46.054899/Model/actor_iteration_999_Player_0.hd5"
loadModelAgent1 =DQLModel0 #""#""#"" #DQLModel #[actorModelA2C,criticModelA2c] #[actorModelDDPG,criticModelDDPG]
loadModelAgent2 =DQLModel4#""#""# ""#[actorModel,criticModel]
loadModelAgent3 =DQLModel9#""#""# ""
loadModelAgent4 =DQLModelr#""#DQLModelr#""#""# ""
#
# loadModel = [loadModelAgent1,loadModelAgent2, loadModelAgent3, loadModelAgent4] #indicate where the saved model is
loadModel = [loadModelAgent3, loadModelAgent4, loadModelAgent1,
loadModelAgent2] # indicate where the saved model is
#
# loadModel = "" #indicate where the saved model is
# #Parameters for controling the experiment
isLogging = False # Logg the experiment
isPlotting = True # plot the experiment
plotFrequency = 1000 # plot the plots every X games
createDataset = True # weather to save the dataset
saveExperimentsIn = "/home/pablo/Documents/Datasets/ChefsHat_ReinforcementLearning/ICPR_Experiments/DQL/Self/1000/" # Directory where the experiment will be saved
winsP1 = []
winsP2 = []
winsP3 = []
winsP4 = []
for i in range(numRuns):
metrics = ChefsHatExperimentHandler.runExperiment(numGames=numGames, playersAgents=playersAgents,experimentDescriptor=experimentDescriptor,isLogging=isLogging,isPlotting=isPlotting,plotFrequency = plotFrequency, createDataset=createDataset,saveExperimentsIn=saveExperimentsIn, loadModel=loadModel, rewardFunction=reward)
# Player1 - agent
p1 = metrics[2]
p2 = metrics[3]
p3 = metrics[4]
p4 = metrics[5]
winsP1.append(p1[0])
winsP2.append(p2[0])
winsP3.append(p3[0])
winsP4.append(p4[0])
print ("Metrics:" + str(metrics))
plotVictoriesTotal(winsP1, winsP2, winsP3, winsP4,numGames, experimentDescriptor, saveExperimentsIn)