def test_get_players_sorted():
"""Run get_players with sorted flag to true."""
# Set up variables
ba1 = BaseAgent()
ba2 = BaseAgent()
cfe = CoinFlipEngine()
lad = BaseLadder(game=cfe)
lad.match_func = mock_match_func
# Add players to the ladder
lad.add_player(ba1)
lad.add_player(ba2)
# Run the game
lad.run_game()
# Check that the results are sorted in ascending elo
players = lad.get_players(sort=True)
player1 = players[0]
player2 = players[1]
assert player1.elo > player2.elo
assert (player1.num_wins == 1 and player2.num_wins == 0)
assert (player1.num_losses == 0 and player2.num_losses == 1)
def test_match_error():
"""Ensure RuntimeError is thrown when no players available to match."""
lad = BaseLadder()
ba1 = BaseAgent(id_in="Ba1")
lad.match_func = mock_match_func
# Error thrown when no players available
try:
lad.match_players()
assert False
except RuntimeError:
pass
# Error thrown when only one player available
lad.add_player(ba1)
try:
lad.match_players()
assert False
except RuntimeError:
pass
# Originally enough players, but not enough afterwards
for _ in range(2):
lad.add_player(BaseAgent())
try:
lad.match_players() # 3 players in pool
lad.match_players() # Only 1 player in pool
assert False
except RuntimeError:
pass
def test_floor():
"""Tests that ratings do not go below 1000."""
player1 = BaseAgent()
player2 = BaseAgent()
player1.elo = elo(player1, player2, 0)
assert player1.elo == 1000
def test_run_game():
"""Test run_game functions properly."""
# Set up variables
ba1 = BaseAgent()
ba2 = BaseAgent()
cfe = CoinFlipEngine()
lad = WeightedLadder(game=cfe)
# Add players to the ladder
lad.add_player(ba1)
lad.add_player(ba2)
# Run the game
lad.run_game()
# Check that the ladder updated properly
players = lad.get_players()
player1 = players[0]
player2 = players[1]
# Only one elo value changes
assert ((player1.elo > 1000 and player2.elo == 1000)
or (player1.elo == 1000 and player2.elo > 1000))
# Someone won the game
assert ((player1.num_wins == 0 and player2.num_wins == 1)
or (player1.num_wins == 1 and player2.num_wins == 0))
# Someone lost the game
assert ((player1.num_losses == 0 and player2.num_losses == 1)
or (player1.num_losses == 1 and player2.num_losses == 0))
def test_make_move():
"""Tests that make_move throws an error."""
ba1 = BaseAgent()
try:
ba1.make_move()
assert False
except NotImplementedError:
return
def test_add():
"""Basic test for ladder add_player method."""
lad = BaseLadder()
ba1 = BaseAgent()
ba2 = BaseAgent()
lad.add_player(ba1)
lad.add_player(ba2)
assert len(lad.player_pool) == 2
def __init__(self):
"""
Init this agent
:param size_x: battlefield horizontal size
:param size_y: battlefield vertical size
:param detector_num: detector quantity of this side
:param fighter_num: fighter quantity of this side
"""
BaseAgent.__init__(self)
self.obs_ind = 'maddpg'
def test_run():
"""Run single iteration of game."""
player1 = BaseAgent()
player2 = BaseAgent()
cfe_win = CoinFlipEngine(prob_win=1)
outcome = cfe_win.run(player1, player2)
assert outcome == 1
cfe_loss = CoinFlipEngine(prob_win=0)
outcome = cfe_loss.run(player1, player2)
assert outcome == 0
def __init__(self):
"""
Init this agent
:param size_x: battlefield horizontal size
:param size_y: battlefield vertical size
:param detector_num: detector quantity of this side
:param fighter_num: fighter quantity of this side
"""
BaseAgent.__init__(self)
self.obs_ind = 'simple'
if not os.path.exists('model/simple/model.pkl'):
print('Error: agent simple model data not exist!')
exit(1)
self.fighter_model = dqn.RLFighter(ACTION_NUM)
def __init__(self):
"""
Init this agent
:param size_x: battlefield horizontal size
:param size_y: battlefield vertical size
:param detector_num: detector quantity of this side
:param fighter_num: fighter quantity of this side
"""
BaseAgent.__init__(self)
self.obs_ind = 'maddpg6'
if not os.path.exists('model/{}/{}0.pkl'.format(MODEL_PATH, MODEL_NAME)):
logger.info('Error: agent maddpg6 model data not exist!')
exit(1)
self.maddpg = MADDPG(FIGHTER_NUM, OBS_NUM, ACTION_NUM)
def __init__(self):
"""
Init this agent
:param size_x: battlefield horizontal size
:param size_y: battlefield vertical size
:param detector_num: detector quantity of this side
:param fighter_num: fighter quantity of this side
"""
BaseAgent.__init__(self)
self.obs_ind = 'airc_homo_rule'
self.last_random_action_set = {}
self.to_explore_set = {}
self.explore_destination_set = {}
self.rule_set = 'attack'
def test_match_basic():
"""Test that match functions properly."""
# Set up variables
lad = WeightedLadder()
ba1 = BaseAgent()
ba2 = BaseAgent()
# Add the players to the ladder
lad.add_player(ba1)
lad.add_player(ba2)
# Generate a match (should be ba1 and ba2)
_ = lad.match_players()
# Assert that players get removed from ladder
assert not lad.player_pool
assert lad.num_turns == 1
def test_init():
"""Test the variables are set properly."""
ba1 = BaseAgent(id_in="doot", type="Pew")
assert ba1.elo == 1000
assert ba1.id == "doot"
assert ba1.num_wins == 0
assert ba1.num_losses == 0
assert ba1.type == "Pew"
def test_match_func():
"""Test the match_func to make sure it works."""
# Set up variables
lad = WeightedLadder()
ba1 = BaseAgent(id_in="Ba1")
ba2 = BaseAgent(id_in="Ba2")
# Make the elo score higher
ba1.elo = 1500
ba2.elo = 1400
# Add the higher ranked players
lad.add_player(ba1)
lad.add_player(ba2)
# Add the rest of the agents (ranked lower)
for i in range(3, 11):
lad.add_player(BaseAgent(id_in="Ba{}".format(i)))
# Try matching the higher ranked players together
match1, match2 = lad.match_players()
while (match1.id is not ba1.id) and (match1.id is not ba2.id):
match1, match2 = lad.match_players()
# Higher elo players got matched together
assert (match2.id == ba1.id or match2.id == ba2.id)
def __init__(self):
"""
Init this agent
:param size_x: battlefield horizontal size
:param size_y: battlefield vertical size
:param detector_num: detector quantity of this side
:param fighter_num: fighter quantity of this side
"""
BaseAgent.__init__(self)
# self.preposs = [[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0],
# [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]
self.obs_ind = OBS_IND_NAME
if not os.path.exists('model/{}/{}.pkl'.format(MODEL_PATH_AGENT,
MODEL_NAME_AGENT)):
logger.info('Error: agent maddpg4 model data not exist!')
exit(1)
self.fighter_model = dqn.RLFighter(OBS_NUM, ACTION_NUM)
self.step = 0
def __init__(self):
"""
Init this agent
:param size_x: battlefield horizontal size
:param size_y: battlefield vertical size
:param detector_num: detector quantity of this side
:param fighter_num: fighter quantity of this side
"""
BaseAgent.__init__(self)
# self.preposs = [[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0],
# [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]
self.obs_ind = 'dqn_jy'
if not os.path.exists('model/dqn_jy/model_000015200.pkl'):
print(
'class Agent(BaseAgent) Error: agent test1 model data not exist!'
)
exit(1)
self.fighter_model = dqn.RLFighter(ACTION_NUM)
def test_available_players():
"""Test that available players picks out players not in games."""
lad = BaseLadder()
ba1 = BaseAgent()
ba2 = BaseAgent()
ba3 = BaseAgent()
lad.match_func = mock_match_func
# No players are in games, so all available
lad.add_player(ba1)
lad.add_player(ba2)
lad.add_player(ba3)
assert len(lad.available_players) == 3
# One player is taken
_, _ = lad.match_players()
assert len(lad.available_players) == 1
assert lad.available_players[0] in [(ba2, 0), (ba3, 0), (ba1, 0)]
def test_drop():
"""Test that ratings drop when appropriate."""
player1 = BaseAgent()
player2 = BaseAgent()
player1.elo = 1619
player2.elo = 1609
player1.elo = elo(player1, player2, 0)
assert player1.elo == 1602
def test_increase():
"""Tests that ratings increase when appropriate."""
player1 = BaseAgent()
player2 = BaseAgent()
player1.elo = 1619
player2.elo = 1609
player2.elo = elo(player2, player1, 1)
assert player2.elo == 1625
def test_duplicate_add():
"""Test that same player cannot exist twice on ladder."""
lad = BaseLadder()
ba1 = BaseAgent()
lad.add_player(ba1)
lad.num_turns = 50
lad.add_player(ba1)
assert lad.player_pool[0][1] == 50
assert len(lad.player_pool) == 1
def test_selection_size():
"""Test that selection size works properly."""
# Initialization
lad = WeightedLadder()
assert lad.selection_size == 1
# Populate ladder
base_player = BaseAgent()
base_player.elo = 1030
for ind in range(15):
new_player = BaseAgent()
new_player.elo = 1000 + ind
lad.add_player(new_player)
# Only getting pool of one player
lad.selection_size = 1
assert len(lad.get_candidate_matches(base_player)) == 1
assert lad.get_candidate_matches(base_player)[0][0].elo == 1014
# Pool of normal size, get five best players
lad.selection_size = 5
assert len(lad.get_candidate_matches(base_player)) == 5
max_elo = 1014
for pair in lad.get_candidate_matches(base_player):
assert pair[0].elo == max_elo
max_elo -= 1
# Pool of unreasonable size, just get max number of players
lad.selection_size = 5000
assert len(lad.get_candidate_matches(base_player)) == 15
def test_match_basic():
"""Test that match functions properly."""
# Set up variables
lad = BaseLadder()
ba1 = BaseAgent()
ba2 = BaseAgent()
# Use fake match function
lad.match_func = mock_match_func
# Add the players to the ladder
lad.add_player(ba1)
lad.add_player(ba2)
# Generate a match (should be ba1 and ba2)
_ = lad.match_players()
# Assert that players get removed from player pool
assert not lad.available_players
assert lad.num_turns == 1
for player, _ in lad.player_pool:
assert player.in_game
def test_no_duplicates():
"""Test that same player cannot exist twice on ladder."""
lad = BaseLadder()
ba1 = BaseAgent()
lad.add_player(ba1)
try:
lad.add_player(ba1)
except ValueError:
# Ladder throws a ValueError if a duplicate player exists
# We want to be here
return
assert False
def test_win_loss():
"""Test calculation of win/loss ratio."""
ba1 = BaseAgent()
# Wins with no losses has undefined w/l ratio
ba1.num_wins = 50
assert ba1.win_loss_ratio() is None
# Losses makes the calculation valid
ba1.num_losses = 10
assert ba1.win_loss_ratio() == 5
# num_wins + num_losses = total_games
assert ba1.total_games() == 60
from agent.base_agent import BaseAgent
import envs
import gym
if __name__ == '__main__':
env = gym.make('2048-v0')
agent = BaseAgent(env)
train_log = agent.train()
test_result = agent.test()
def add_agents(self):
"""Add agents to the ladder."""
for _ in range(self.num_players):
player = BaseAgent()
self.ladder.add_player(player)
def run_agent(agent: BaseAgent, reward: dict, obs: dict):
action, event_type = agent.step(reward, obs)
reward = play_ground.act(action, event_type) # reward after an action
return reward
