def test_get_mean_and_standard_deviation_difference_results():
"""Tests that get_mean_and_standard_deviation_difference_results method produces correct output"""
results = [[1.0, 2.0, 3.0], [5.0, -33.0, 55.0], [2.5, 2.5, 2.5]]
mean_results = [
np.mean([1.0, 5.0, 2.5]),
np.mean([2.0, -33.0, 2.5]),
np.mean([3.0, 55.0, 2.5])
]
std_results = [
np.std([1.0, 5.0, 2.5]),
np.std([2.0, -33.0, 2.5]),
np.std([3.0, 55.0, 2.5])
]
mean_minus_1_std = [
mean - std_val for mean, std_val in zip(mean_results, std_results)
]
mean_plus_1_std = [
mean + std_val for mean, std_val in zip(mean_results, std_results)
]
config = Config()
config.standard_deviation_results = 1.0
trainer = Trainer(config, [])
mean_minus_x_std_guess, mean_results_guess, mean_plus_x_std_guess = trainer.get_mean_and_standard_deviation_difference_results(
results)
assert mean_results == mean_results_guess
assert mean_minus_1_std == mean_minus_x_std_guess
assert mean_plus_1_std == mean_plus_x_std_guess
config.standard_deviation_results = 3.0
trainer = Trainer(config, [])
mean_minus_x_std_guess, mean_results_guess, mean_plus_x_std_guess = trainer.get_mean_and_standard_deviation_difference_results(
results)
mean_plus_3_std = [
mean + 3.0 * std_val
for mean, std_val in zip(mean_results, std_results)
]
mean_minus_3_std = [
mean - 3.0 * std_val
for mean, std_val in zip(mean_results, std_results)
]
assert mean_results == mean_results_guess
assert mean_minus_3_std == mean_minus_x_std_guess
assert mean_plus_3_std == mean_plus_x_std_guess
import gym, os
from agents.hierarchical_agents.DIAYN import DIAYN
from agents.hierarchical_agents.DBH import DBH
from agents.actor_critic_agents.SAC_Discrete import SAC_Discrete
from agents.actor_critic_agents.SAC import SAC
from agents.DQN_agents.DDQN import DDQN
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
import argparse
config = Config()
parser = argparse.ArgumentParser()
parser.add_argument('--env',
action='store',
dest='environment',
default='SpaceInvaders-v0',
help='which environment to compare on')
parser.add_argument('--alg',
nargs='+',
action='store',
dest='algorithms',
default='SAC_Discrete',
help='which algorithms to compare')
parser.add_argument('--eval',
type=bool,
default=False,
action='store',
dest='evaluate',
help='set False for training and True for evaluating.')
parser.add_argument('--num_ep',
type=int,
import os
import sys
from os.path import dirname, abspath
sys.path.append(dirname(dirname(abspath(__file__))))
import gym
from agents.actor_critic_agents.SAC_Discrete import SAC_Discrete
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
from environments.DMP_Env_1D_dynamic import deep_mobile_printing_1d1r
config = Config()
config.seed = 1
config.environment = deep_mobile_printing_1d1r()
config.num_episodes_to_run = 10000
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 = True
config.GPU = "cuda:0"
config.overwrite_existing_results_file = True
config.randomise_random_seed = False
config.save_model = False
OUT_FILE_NAME = "SAC_1d" + "sin" + "_seed_" + str(config.seed)
config.save_model_path = "/mnt/NAS/home/WenyuHan/SNAC/SAC/1D/dynamic/" + OUT_FILE_NAME + "/"
config.file_to_save_data_results = "/mnt/NAS/home/WenyuHan/SNAC/SAC/1D/dynamic/" + OUT_FILE_NAME + "/" + "Results_Data.pkl"
config.file_to_save_results_graph = "/mnt/NAS/home/WenyuHan/SNAC/SAC/1D/dynamic/" + OUT_FILE_NAME + "/" + "Results_Graph.png"
if os.path.exists(config.save_model_path) == False:
os.makedirs(config.save_model_path)
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.DIAYN import DIAYN
from hierarchical_agents.HRL import HRL
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 = 200
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
from gym.wrappers import FlattenDictWrapper
from agents.DQN_agents.DQN_HER import DQN_HER
from environments.Bit_Flipping_Environment import Bit_Flipping_Environment
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 = Bit_Flipping_Environment(14)
config.num_episodes_to_run = 4500
config.file_to_save_data_results = None #"Data_and_Graphs/Bit_Flipping_Results_Data.pkl"
config.file_to_save_results_graph = None #"Data_and_Graphs/Bit_Flipping_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 = {
"DQN_Agents": {
"learning_rate": 0.001,
"batch_size": 128,
"buffer_size": 100000,
"epsilon_decay_rate_denominator": 150,
"discount_rate": 0.999,
"incremental_td_error": 1e-8,
import pickle
from Trainer import Trainer
from utilities.data_structures.Config import Config
trainer = Trainer(config=Config(), agents=None)
#
# trainer.visualise_set_of_preexisting_results(save_image_path="Four_Rooms_and_Long_Corridor.png", results_data_paths=["Long_Corridor_Results_Data.pkl", "Four_Rooms.pkl"],
# plot_titles=["Long Corridor", "Four Rooms"], y_limits=[(0.0, 0.25), (-90.0, 100.25)])
trainer.visualise_preexisting_results(
save_image_path="hrl_experiments/Taxi_graph_comparison.png",
data_path="hrl_experiments/Taxi_data.pkl",
title="Taxi v2",
y_limits=(-800.0, 0.0))
# trainer.visualise_preexisting_results(save_image_path="Long_Corridor_Graph.png", data_path="Long_Corridor_Results_Data.pkl",
# title="Long Corridor", y_limits=(0.0, 0.25))
# trainer.visualise_preexisting_results(save_image_path="Hopper_Results_Graph_Both_Agents.png", data_path="Hopper_Results_Data.pkl",
# title="Hopper") #, y_limits=(0.0, 0.25))
# trainer.visualise_set_of_preexisting_results(results_data_paths=["Cart_Pole_Results_Data.pkl",
# "Mountain_Car_Results_Data.pkl"],
# plot_titles=["Cart Pole (Discrete Actions)", "Mountain Car (Continuous Actions)"],
# save_image_path="CartPole_and_MountainCar_Graph.png")
# trainer.visualise_set_of_preexisting_results(results_data_paths=["Data_and_Graphs/Bit_Flipping_Results_Data.pkl",
# "Data_and_Graphs/Fetch_Reach_Results_Data.pkl"],
# plot_titles=["Bit Flipping", "Fetch Reach"],
import gym
import pytest
from utilities.Utility_Functions import flatten_action_id_to_actions
from utilities.data_structures.Config import Config
config = Config()
config.seed = 1
config.environment = gym.make("Taxi-v2")
config.env_parameters = {}
config.num_episodes_to_run = 1000
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.load_model = False
linear_hidden_units = [10, 5]
learning_rate = 0.01
buffer_size = 40000
batch_size = 256
batch_norm = False
embedding_dimensionality = 15
gradient_clipping_norm = 5
import gym
from A2C import A2C
from Dueling_DDQN import Dueling_DDQN
from SAC_Discrete import SAC_Discrete
from agents.actor_critic_agents.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 = 450
config.file_to_save_data_results = "data_and_graphs/Cart_Pole_Results_Data.pkl"
config.file_to_save_results_graph = "data_and_graphs/Cart_Pole_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 = 1
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
import os
import sys
sys.path.append(os.path.dirname(os.path.abspath(os.path.dirname(__file__))))
import gym
from agents.actor_critic_agents.DDPG import DDPG
from agents.actor_critic_agents.DDPG_HER_Che import DDPG_HER_Che
from utilities.data_structures.Config import Config
from agents.Trainer import Trainer
config = Config()
config.seed = 1
config.environment = gym.make("FetchReach-v1")
#config.environment = gym.make("FetchPush-v1")
#config.environment = gym.make("FetchPickAndPlace-v1")
#config.environment = gym.make("FetchSlide-v1")
config.num_episodes_to_run = 2000
#config.num_episodes_to_run = 2000
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.runs_per_agent = 25
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
import sys sys.path.insert(0, '../') from environments.isc_environments.SimpleISC import SimpleISC from utilities.data_structures.Config import Config from agents.Trainer import Trainer from agents.actor_critic_agents import A2C, A3C, DDPG, DDPG_HER from gym.core import Wrapper from torch.cuda import is_available config = Config() config.environment = Wrapper(SimpleISC(mode="DISCRETE")) config.num_episodes_to_run = 5 config.file_to_save_data_results = "results/data_and_graphs/isc/IllinoisSolarCar_Results_Data.pkl" config.file_to_save_results_graph = "results/data_and_graphs/isc/IllinoisSolarCar_Results_Graph.png" config.show_solution_score = False config.visualise_individual_results = False config.visualise_overall_agent_results = False config.standard_deviation_results = 1.0 config.runs_per_agent = 1 config.use_GPU = is_available() config.overwrite_existing_results_file = True config.randomise_random_seed = False config.save_model = False config.seed = 0 config.debug_mode = True config.wandb_log = True
from agents.DQN_agents.DDQN import DDQN
from agents.actor_critic_agents.DDPG import DDPG
from agents.actor_critic_agents.SAC import SAC
from agents.actor_critic_agents.SAC_Discrete import SAC_Discrete
from agents.actor_critic_agents.A3C import A3C
from agents.DQN_agents.DDQN import DDQN
from agents.DQN_agents.Dueling_DDQN import Dueling_DDQN
from environments.VEC_Environment import VEC_Environment
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
import matplotlib.pyplot as plt
import numpy as np
config = Config()
config.seed = 1
config.num_episodes_to_run = 8000
# config.file_to_save_data_results = "results/data_and_graphs/VEC.pkl"
# config.file_to_save_results_graph = "results/data_and_graphs/VEC.png"
config.show_solution_score = False
config.visualise_individual_results = False
config.visualise_overall_agent_results = False
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = True
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
config.device = "cuda:0"
config.hyperparameters = {
import gym
from HRL import HRL
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 = {
"HRL": {
"linear_hidden_units": [10, 5],
"learning_rate":
0.01,
from environments.isc_environments.SimpleISC import SimpleISC from utilities.data_structures.Config import Config from agents.Trainer import Trainer from agents.DQN_agents import DQN, DDQN, Dueling_DDQN, DDQN_With_Prioritised_Experience_Replay, DRQN import wandb from gym.core import Wrapper from torch.cuda import is_available config = Config() config.environment = Wrapper(SimpleISC(mode="DISCRETE")) config.num_episodes_to_run = 5_000 config.file_to_save_data_results = "results/data_and_graphs/isc/IllinoisSolarCar_Results_Data.pkl" config.runs_per_agent = 1 config.use_GPU = is_available() config.overwrite_existing_results_file = True config.randomise_random_seed = False config.save_model = False config.model = None config.seed = 0 config.debug_mode = True config.wandb_log = True config.wandb_job_type = "testing" config.wandb_entity = "rafael_piacsek" config.wandb_tags = ["initial testing"] config.wandb_model_log_freq = 1_000
import gym
from agents.hierarchical_agents.DIAYN import DIAYN
from agents.actor_critic_agents.SAC import SAC
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
config = Config()
config.environment = gym.make("MountainCarContinuous-v0")
config.seed = 1
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 = True
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
linear_hidden_units = [32, 32]
learning_rate = 0.01
buffer_size = 100000
batch_size = 256
batch_norm = False
embedding_dimensionality = 10
gradient_clipping_norm = 5
update_every_n_steps = 1
"""Tests for the hierarchical RL agent HIRO"""
import copy
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
from agents.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.load_model = False
config.hyperparameters = {
"h_DQN": {
"CONTROLLER": {
"batch_size":
# import the agents and the trainer
from agents.DQN_agents.DQN_multi_agent import DQN
from agents.Trainer_multi_agent import Trainer
routing_modes = [
"Q_routing_2_hop", "Q_routing_1_hop", "Q_routing_0_hop", "Q_routing",
"TTSPWRR", "TTSP"
]
network_names = ["5x6", "UES_Manhatan", "toronto"]
gpu_num = int(sys.argv[1])
algorithm_num = int(sys.argv[2])
network_num = int(sys.argv[3])
config = Config()
config.use_GPU = True
assert (torch.cuda.is_available())
config.device = torch.device(gpu_num)
config.routing_mode = routing_modes[algorithm_num]
network_name = network_names[network_num]
config.training_mode = True
config.does_need_network_state = config.routing_mode in [
"Q_routing_2_hop", "Q_routing_1_hop", "Q_routing_0_hop"
]
config.does_need_network_state_embeding = config.routing_mode in [
"Q_routing_2_hop", "Q_routing_1_hop"
from environments.j2n6s300.DDPG_HER_env_Gazebo import j2n6s300_Environment
from agents.actor_critic_agents.DDPG_HER import DDPG_HER
from utilities.data_structures.Config import Config
from agents.Trainer import Trainer
from datetime import datetime
import os
now = datetime.now() # current date and time
date_str = now.strftime("%Y-%m-%d_%H-%M-%S")
os.mkdir('Data_and_Graphs/results_' + date_str)
path = 'Data_and_Graphs/results_' + date_str + '/'
config = Config()
config.seed = 1
config.environment = j2n6s300_Environment(proxyID='Env1')
config.num_episodes_to_run = 1
config.file_to_save_config = path + "config.json"
config.file_to_save_data_results = path + "jaco_DDPG-HER.pkl"
config.file_to_save_results_graph = path + "jaco_DDPG-HER.png"
config.show_solution_score = False
config.visualise_results_while_training = True
config.visualise_individual_results = True
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = True
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.load_model = False
config.load_model_path = "Models/model.pt"
config.save_model = True
from environments.isc_environments.SimpleISC import SimpleISC from utilities.data_structures.Config import Config from agents.Trainer import Trainer from agents.DQN_agents.DQN import DQN from agents.DQN_agents.DDQN import DDQN from agents.DQN_agents.Dueling_DDQN import Dueling_DDQN from agents.DQN_agents.DDQN_With_Prioritised_Experience_Replay import DDQN_With_Prioritised_Experience_Replay from agents.DQN_agents.DRQN import DRQN from models.FCNN import FCNN from gym.core import Wrapper config = Config() config.environment = Wrapper(SimpleISC(mode="DISCRETE")) config.num_episodes_to_run = 50 config.file_to_save_data_results = "results/data_and_graphs/isc/IllinoisSolarCar_Results_Data.pkl" config.file_to_save_results_graph = "results/data_and_graphs/isc/IllinoisSolarCar_Results_Graph.png" config.show_solution_score = True config.visualise_individual_results = True 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 = True config.randomise_random_seed = False config.save_model = False
import gym
from agents.policy_gradient_agents.PPO import PPO
from agents.actor_critic_agents.DDPG import DDPG
from agents.actor_critic_agents.SAC import SAC
from agents.actor_critic_agents.TD3 import TD3
from agents.Trainer import Trainer
from agents.hierarchical_agents.DIAYN import DIAYN
from utilities.data_structures.Config import Config
config = Config()
config.seed = 1
config.environment = gym.make("Hopper-v2")
config.num_episodes_to_run = 1000
config.file_to_save_data_results = "data_and_graphs/Hopper_Results_Data.pkl"
config.file_to_save_results_graph = "data_and_graphs/Hopper_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.load_model = False
actor_critic_agent_hyperparameters = {
"Actor": {
"learning_rate": 0.0003,
import gym
from agents.hierarchical_agents.HRL.HRL import HRL
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
config = Config()
config.environment = gym.make("Taxi-v2")
config.seed = 1
config.env_parameters = {}
config.num_episodes_to_run = 2000
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
linear_hidden_units = [32, 32]
learning_rate = 0.01
buffer_size = 100000
batch_size = 256
batch_norm = False
embedding_dimensionality = 10
gradient_clipping_norm = 5
update_every_n_steps = 1
from agents.actor_critic_agents.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
from agents.policy_gradient_agents.REINFORCE import REINFORCE
## envs import ##
from environments.carla_enviroments import env_v1_ObstacleAvoidance
env_title = "ObstacleAvoidance-v0"
config = Config()
config.env_title = env_title
config.seed = 1
config.environment = gym.make(env_title)
config.num_episodes_to_run = 2000
config.show_solution_score = False
config.visualise_individual_results = True
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = True
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = True
config.log_loss = False
config.log_base = time.strftime("%Y%m%d%H%M%S", time.localtime())
from agents.DQN_agents.DQN_HER import DQN_HER
from agents.DQN_agents.DQN_With_Fixed_Q_Targets import DQN_With_Fixed_Q_Targets
from agents.DQN_agents.Dueling_DDQN import Dueling_DDQN
from agents.DQN_agents.DDQN import DDQN
from agents.hierarchical_agents.DIAYN import DIAYN
from agents.hierarchical_agents.h_DQN import h_DQN
from agents.hierarchical_agents.HIRO import HIRO
from agents.hierarchical_agents.SNN_HRL import SNN_HRL
from agents.policy_gradient_agents.PPO import PPO
from agents.policy_gradient_agents.REINFORCE import REINFORCE
from environments.FaceDiscreete import FaceEnvironementDiscreete
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
config = Config()
config.seed = 1
config.environment = FaceEnvironementDiscreete(
"../weights/blg_small_12_5e-06_5e-05_2_8_small_big_noisy_first_True_512")
config.num_episodes_to_run = 500
config.file_to_save_data_results = "Data_and_Graphs/FaceDiscreete.pkl"
config.file_to_save_results_graph = "Data_and_Graphs/FaceDiscreete.png"
config.show_solution_score = True
config.visualise_individual_results = True
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = True
config.overwrite_existing_results_file = True
import gym
from models.Trainer import Trainer
from models.actor_critic_agents.DDPG import DDPG
from models.actor_critic_agents.TD3 import TD3
from models.policy_gradient_agents.PPO import PPO
from utilities.data_structures.Config import Config
config = Config()
config.seed = 1
config.environment = gym.make("MountainCarContinuous-v0")
config.num_episodes_to_run = 450
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",
"learning_iterations_per_round": 10,
"discount_rate": 0.9,
self.logger.info("Learning rate {}".format(new_lr))
if __name__ == '__main__':
from utilities.data_structures.Config import Config
import gym
## envs import ##
from environments.carla_enviroments import env_v1_ObstacleAvoidance
# net = q_network_toa(n_action=4)
# net.to('cuda')
# input = torch.rand(size=(10, 3, 224, 224)).to('cuda')
# q1, q2 = net(input)
config = Config()
config.seed = 1
config.environment = gym.make("ObstacleAvoidance-v0")
config.num_episodes_to_run = 2000
config.file_to_save_data_results = "C:/my_project/Deep-Reinforcement-Learning-Algorithms-with-PyTorch/results/data_and_graphs/carla_obstacle_avoidance/data.pkl"
config.file_to_save_results_graph = "C:/my_project/Deep-Reinforcement-Learning-Algorithms-with-PyTorch/results/data_and_graphs/carla_obstacle_avoidance/data.png"
config.show_solution_score = False
config.visualise_individual_results = True
config.visualise_overall_agent_results = True
config.standard_deviation_results = 1.0
config.runs_per_agent = 1
config.use_GPU = True
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = True
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_With_Fixed_Q_Targets import DQN_With_Fixed_Q_Targets
from environments.Bit_Flipping_Environment import Bit_Flipping_Environment
from agents.policy_gradient_agents.PPO import PPO
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
from agents.DQN_agents.DQN import DQN
import numpy as np
import torch
random.seed(1)
np.random.seed(1)
torch.manual_seed(1)
config = Config()
config.seed = 1
config.environment = Bit_Flipping_Environment(4)
config.num_episodes_to_run = 1
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.visualise_individual_results = False
config.visualise_overall_agent_results = False
config.randomise_random_seed = False
config.runs_per_agent = 1
config.use_GPU = False
config.load_model = False
config.hyperparameters = {
"DQN_Agents": {
"learning_rate": 0.005,
"batch_size": 3,
import gym
from agents.Trainer import Trainer
from agents.actor_critic_agents.DDPG import DDPG
from agents.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.load_model = False
config.hyperparameters = {
"HIRO": {
"LOWER_LEVEL": {
"max_lower_level_timesteps": 5,
"Actor": {
"learning_rate": 0.001,
"linear_hidden_units": [20, 20],
"final_layer_activation": "TANH",
import gym
from environments.Atari_Environment import make_atari_game
from agents.DQN_agents.DDQN import DDQN
from agents.hierarchical_agents.HRL.HRL import HRL
from agents.hierarchical_agents.HRL.Model_HRL import Model_HRL
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
config = Config()
config.seed = 1
config.environment = make_atari_game("SpaceInvaders-v0")
config.env_parameters = {}
config.num_episodes_to_run = 500
config.file_to_save_data_results = "data_and_graphs/hrl_experiments/Space_Invaders_Data.pkl"
config.file_to_save_results_graph = "data_and_graphs/hrl_experiments/Space_Invaders.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 = 10
config.use_GPU = False
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.save_model = False
# Loss is not drawing a random sample! otherwise wouldnt jump around that much!!
linear_hidden_units = [32, 32]
learning_rate = 0.005 # 0.001 taxi
buffer_size = 1000000
batch_size = 256
# from environments.Bit_Flipping_Environment import Bit_Flipping_Environment
# from agents.policy_gradient_agents.PPO import PPO
# from environments.Four_Rooms_Environment import Four_Rooms_Environment
# from agents.hierarchical_agents.SNN_HRL import SNN_HRL
# from agents.actor_critic_agents.TD3 import TD3
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
from agents.DQN_agents.DQN import DQN
import numpy as np
import torch
random.seed(1)
np.random.seed(1)
torch.manual_seed(1)
config = Config()
config.seed = 1
config.environment = Bit_Flipping_Environment(4)
config.num_episodes_to_run = 2000
config.file_to_save_data_results = None
config.file_to_save_results_graph = None
config.visualise_individual_results = False
config.visualise_overall_agent_results = False
config.randomise_random_seed = False
config.runs_per_agent = 1
config.use_GPU = False
config.hyperparameters = {
"DQN_Agents": {
"learning_rate": 0.005,
"batch_size": 64,
"buffer_size": 40000,
@author: kashishg
"""
from agents.DQN_agents.DQN_HER import DQN_HER
from environments.j2n6s300.HER_env_tf import j2n6s300_Environment
from agents.Trainer import Trainer
from utilities.data_structures.Config import Config
from datetime import datetime
now = datetime.now() # current date and time
num_episodes_to_run = 500
eps_decay_rate_denom = round(num_episodes_to_run/6)
config = Config()
config.seed = 1
config.environment = j2n6s300_Environment()
config.num_episodes_to_run = num_episodes_to_run
config.file_to_save_data_results = "Data_and_Graphs/{}jaco.pkl".format(now.strftime("%Y-%m-%d_%H-%M-%S_"))
config.file_to_save_results_graph = "Data_and_Graphs/{}jaco.png".format(now.strftime("%Y-%m-%d_%H-%M-%S_"))
config.show_solution_score = False
config.visualise_results_while_training = True
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 = True
config.overwrite_existing_results_file = False
config.randomise_random_seed = True
config.load_model = False
