import gym
import random
import numpy as np
import torch
from Hierarchical_Agents.HIRO import HIRO
from Utilities.Data_Structures.Config import Config
random.seed(1)
np.random.seed(1)
torch.manual_seed(1)
config = Config()
config.seed = 1
config.environment = gym.make("Pendulum-v0")
config.num_episodes_to_run = 1500
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.hyperparameters = {
"LOWER_LEVEL": {
import gym
from Agents.Trainer import Trainer
from DDPG import DDPG
from Hierarchical_Agents.HIRO import HIRO
from Utilities.Data_Structures.Config import Config
config = Config()
config.seed = 1
config.environment = gym.make(
"Reacher-v2") # Reacher-v2 "InvertedPendulum-v2") #Pendulum-v0
config.num_episodes_to_run = 1500
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.hyperparameters = {
"HIRO": {
"LOWER_LEVEL": {
"max_lower_level_timesteps": 5,
from Hierarchical_Agents.SNN_HRL import SNN_HRL
from Agents.Trainer import Trainer
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DQN import DQN
from Agents.Hierarchical_Agents.h_DQN import h_DQN
from Environments.Long_Corridor_Environment import Long_Corridor_Environment
config = Config()
config.seed = 1
config.env_parameters = {"stochasticity_of_action_right": 0.5}
config.environment = Long_Corridor_Environment(
stochasticity_of_action_right=config.
env_parameters["stochasticity_of_action_right"])
config.num_episodes_to_run = 10000
config.file_to_save_data_results = "Data_and_Graphs/Long_Corridor_Results_Data.pkl"
config.file_to_save_results_graph = "Data_and_Graphs/Long_Corridor_Results_Graph.png"
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 3
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.hyperparameters = {
"h_DQN": {
"CONTROLLER": {
"batch_size":
256,
import gym
from A3C import A3C
from Agents.Policy_Gradient_Agents.PPO import PPO
from Agents.Trainer import Trainer
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DDQN import DDQN
from Agents.DQN_Agents.DDQN_With_Prioritised_Experience_Replay import DDQN_With_Prioritised_Experience_Replay
from Agents.DQN_Agents.DQN import DQN
from Agents.DQN_Agents.DQN_With_Fixed_Q_Targets import DQN_With_Fixed_Q_Targets
config = Config()
config.seed = 1
config.environment = gym.make("CartPole-v0")
config.num_episodes_to_run = 1500
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.hyperparameters = {
"DQN_Agents": {
"learning_rate": 0.005,
"batch_size": 128,
from Agents.DQN_Agents.DQN_HER_Agent import DQN_HER_Agent
from Environments.Other_Enrivonments.Bit_Flipping_Environment import Bit_Flipping_Environment
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DQN_Agent import DQN_Agent
from Utilities.Utility_Functions import run_games_for_agents
config = Config()
config.seed = 100
config.environment = Bit_Flipping_Environment(14)
config.max_episodes_to_run = 6000
config.file_to_save_data_results = "Results_Data.pkl"
config.file_to_save_data_results_graph = "Results_Graph.png"
config.visualise_individual_results = True
config.visualise_overall_results = True
config.runs_per_agent = 3
config.use_GPU = False
config.hyperparameters = {
"DQN_Agents": {
"learning_rate": 0.001,
"batch_size": 128,
"buffer_size": 100000,
"epsilon": 0.1,
"epsilon_decay_rate_denominator": 500,
"discount_rate": 0.98,
"tau": 0.1,
"alpha_prioritised_replay": 0.6,
"beta_prioritised_replay": 0.4,
"incremental_td_error": 1e-8,
"update_every_n_steps": 1,
"nn_layers": 2,
import gym
from Agents.Policy_Gradient_Agents.PPO import PPO
from Agents.Actor_Critic_Agents.DDPG import DDPG
from TD3 import TD3
from Agents.Trainer import Trainer
from Utilities.Data_Structures.Config import Config
config = Config()
config.seed = 1
config.environment = gym.make("MountainCarContinuous-v0")
config.num_episodes_to_run = 600
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 3
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.hyperparameters = {
"Policy_Gradient_Agents": {
"learning_rate": 0.05,
"linear_hidden_units": [30, 15],
"final_layer_activation": "TANH",
from A3C import A3C
from Agents.DQN_Agents.DQN_HER import DQN_HER
from DDQN import DDQN
from Environments.Four_Rooms_Environment import Four_Rooms_Environment
from Hierarchical_Agents.SNN_HRL import SNN_HRL
from Agents.Trainer import Trainer
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DQN import DQN
config = Config()
config.seed = 1
config.environment = Four_Rooms_Environment(
15,
15,
stochastic_actions_probability=0.25,
random_start_user_place=True,
random_goal_place=False)
config.num_episodes_to_run = 1000
config.file_to_save_data_results = "Data_and_Graphs/Four_Rooms.pkl"
config.file_to_save_results_graph = "Data_and_Graphs/Four_Rooms.png"
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 3
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
from Agents.Trainer import Trainer
from DDPG import DDPG
from Hierarchical_Agents.HIRO import HIRO
from Utilities.Data_Structures.Config import Config
import gym
config = Config()
config.seed = 1
config.environment = gym.make("Hopper-v2")
config.num_episodes_to_run = 1500
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
DDPG_hyperparameters = { # hyperparameters taken from https://arxiv.org/pdf/1802.09477.pdf
"Actor": {
"learning_rate": 0.001,
"linear_hidden_units": [30, 20],
"final_layer_activation": "TANH",
"batch_norm": False,
"tau": 0.01,
"gradient_clipping_norm": 5
},
import gym
from Hierarchical_Agents.SNN_HRL import SNN_HRL
from Agents.Trainer import Trainer
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DQN import DQN
from Agents.Hierarchical_Agents.h_DQN import h_DQN
config = Config()
config.seed = 1
config.environment = gym.make("Taxi-v2")
config.env_parameters = {}
config.num_episodes_to_run = 10000
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 3
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.hyperparameters = {
"h_DQN": {
"CONTROLLER": {
"batch_size":
256,
"learning_rate":
from Agents.Policy_Gradient_Agents.PPO_Agent import PPO_Agent
from Utilities.Data_Structures.Config import Config
from Agents.DQN_Agents.DDQN_Agent import DDQN_Agent
from Agents.DQN_Agents.DDQN_With_Prioritised_Experience_Replay import DDQN_With_Prioritised_Experience_Replay
from Agents.DQN_Agents.DQN_Agent import DQN_Agent
from Agents.DQN_Agents.DQN_Agent_With_Fixed_Q_Targets import DQN_Agent_With_Fixed_Q_Targets
from Environments.Open_AI_Gym_Environments.Cart_Pole_Environment import Cart_Pole_Environment
from Agents.Policy_Gradient_Agents.REINFORCE_Agent import REINFORCE_Agent
from Agents.Stochastic_Policy_Search_Agents.Genetic_Agent import Genetic_Agent
from Agents.Stochastic_Policy_Search_Agents.Hill_Climbing_Agent import Hill_Climbing_Agent
from Utilities.Utility_Functions import run_games_for_agents
config = Config()
config.seed = 1
config.environment = Cart_Pole_Environment()
config.max_episodes_to_run = 2000
config.file_to_save_data_results = "Results_Data.pkl"
config.file_to_save_data_results_graph = "Results_Graph.png"
config.visualise_individual_results = True
config.visualise_overall_results = True
config.runs_per_agent = 1
config.use_GPU = False
config.hyperparameters = {
"DQN_Agents": {
"learning_rate": 0.005,
"batch_size": 256,
"buffer_size": 40000,
"epsilon": 0.1,
"epsilon_decay_rate_denominator": 200,
"discount_rate": 0.99,