def test_train(self):
norm_step = 1100
memory_init_size = 100
step_num = 1500
sess = tf.InteractiveSession()
tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(sess=sess,
scope='dqn',
replay_memory_size = 500,
replay_memory_init_size=memory_init_size,
update_target_estimator_every=100,
norm_step=norm_step,
state_shape=[2],
mlp_layers=[10,10])
sess.run(tf.global_variables_initializer())
predicted_action = agent.eval_step({'obs': np.random.random_sample((2,)), 'legal_actions': [0, 1]})
self.assertGreaterEqual(predicted_action, 0)
self.assertLessEqual(predicted_action, 1)
for step in range(step_num):
ts = [{'obs': np.random.random_sample((2,)), 'legal_actions': [0, 1]}, np.random.randint(2), 0, {'obs': np.random.random_sample((2,)), 'legal_actions': [0, 1]}, True]
agent.feed(ts)
if step > norm_step + memory_init_size:
agent.train()
predicted_action = agent.step({'obs': np.random.random_sample((2,)), 'legal_actions': [0, 1]})
self.assertGreaterEqual(predicted_action, 0)
self.assertLessEqual(predicted_action, 1)
sess.close()
tf.reset_default_graph()
def __init__(self, graph, sess):
""" Load pretrained model
"""
super().__init__()
self.graph = graph
self.sess = sess
env = rlcard.make('tarot')
with self.graph.as_default():
self.dqn_agent = DQNAgent(
self.sess,
scope='dqn',
action_num=78, # env.action_num,
replay_memory_size=20000,
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[512, 1024, 512])
normalize(env, self.dqn_agent, 1000)
self.sess.run(tf.compat.v1.global_variables_initializer())
check_point_path = os.path.join(ROOT_PATH, 'tarot_v180007')
with self.sess.as_default():
with self.graph.as_default():
saver = tf.compat.v1.train.Saver(
tf.compat.v1.model_variables())
saver.restore(self.sess,
tf.train.latest_checkpoint(check_point_path))
def __init__(self):
''' Load pretrained model
'''
import tensorflow as tf
self.graph = tf.Graph()
self.sess = tf.Session(graph=self.graph)
env = YanivEnv({})
with self.graph.as_default():
self.dqn_agents = []
agent = DQNAgent(self.sess,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=1000,
train_every=1,
state_shape=env.state_shape,
mlp_layers=[512, 512]
)
self.dqn_agents.append(agent)
with self.sess.as_default():
with self.graph.as_default():
saver = tf.train.Saver()
saver.restore(self.sess, tf.train.latest_checkpoint('models/yaniv_dqn'))
def test_init(self):
sess = tf.compat.v1.InteractiveSession()
tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(sess=sess,
scope='dqn',
replay_memory_size=0,
replay_memory_init_size=0,
update_target_estimator_every=0,
discount_factor=0,
epsilon_start=0,
epsilon_end=0,
epsilon_decay_steps=0,
batch_size=0,
action_num=2,
state_shape=[1],
mlp_layers=[10, 10])
self.assertEqual(agent.replay_memory_init_size, 0)
self.assertEqual(agent.update_target_estimator_every, 0)
self.assertEqual(agent.discount_factor, 0)
self.assertEqual(agent.epsilon_decay_steps, 0)
self.assertEqual(agent.batch_size, 0)
self.assertEqual(agent.action_num, 2)
sess.close()
tf.compat.v1.reset_default_graph()
def load_dqn_leduc_agent(model_path):
# Set a global seed
set_global_seed(0)
# Load pretrained model
# tf.reset_default_graph()
graph = tf.Graph()
sess = tf.Session(graph=graph)
with graph.as_default():
nfsp_agents = []
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
train_every=train_every,
state_shape=env.state_shape,
mlp_layers=[128, 128])
# We have a pretrained model here. Change the path for your model.
# check_point_path = os.path.join(rlcard.__path__[0], 'models/pretrained/leduc_holdem_nfsp')
check_point_path = model_path
with sess.as_default():
with graph.as_default():
saver = tf.train.Saver()
saver.restore(sess, tf.train.latest_checkpoint(check_point_path))
return agent
def run(self):
import tensorflow as tf
self.env = rlcard.make('leduc-holdem')
self.sess = tf.Session()
agent = DQNAgent(self.sess,
scope='sub-dqn' + str(self.index),
action_num=self.env.action_num,
replay_memory_init_size=memory_init_size,
train_every=train_every,
state_shape=self.env.state_shape,
mlp_layers=[128, 128])
random_agent = RandomAgent(action_num=self.env.action_num)
self.env.set_agents([agent, random_agent])
self.sess.run(tf.global_variables_initializer())
# Receive instruction to run game and generate trajectories
while True:
instruction = self.input_queue.get()
if instruction is not None:
tasks, train_flag, variables, total_t = instruction
# For evaluation
if not train_flag:
agent.total_t = total_t
global_vars = [
tf.convert_to_tensor(var) for var in variables
]
agent.copy_params_op(global_vars)
for _ in range(tasks):
_, payoffs = self.env.run(is_training=train_flag)
self.output_queue.put(payoffs)
# For training
else:
for _ in range(tasks):
trajectories, _ = self.env.run(is_training=train_flag)
self.output_queue.put(
(trajectories, self.env.timestep))
self.input_queue.task_done()
else:
self.input_queue.task_done()
break
self.sess.close()
return
def __init__(self):
env = rlcard.make('uno')
self.sess1 = tf.compat.v1.Session()
global_step = tf.Variable(0, name='global_step', trainable=False)
self.agent = DQNAgent(self.sess1,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[100, 100])
self.sess1.run(tf.global_variables_initializer())
self.saver = tf.train.Saver()
self.saver.restore(self.sess1,
'./experiments/uno_dqn_result/models/model1.ckpt')
def test_init(self):
agent = DQNAgent(replay_memory_size=0,
replay_memory_init_size=0,
update_target_estimator_every=0,
discount_factor=0,
epsilon_start=0,
epsilon_end=0,
epsilon_decay_steps=0,
batch_size=0,
num_actions=2,
state_shape=[1],
mlp_layers=[10,10],
device=torch.device('cpu'))
self.assertEqual(agent.replay_memory_init_size, 0)
self.assertEqual(agent.update_target_estimator_every, 0)
self.assertEqual(agent.discount_factor, 0)
self.assertEqual(agent.epsilon_decay_steps, 0)
self.assertEqual(agent.batch_size, 0)
self.assertEqual(agent.num_actions, 2)
def test_train(self):
memory_init_size = 100
num_steps = 500
agent = DQNAgent(replay_memory_size = 200,
replay_memory_init_size=memory_init_size,
update_target_estimator_every=100,
state_shape=[2],
mlp_layers=[10,10],
device=torch.device('cpu'))
predicted_action, _ = agent.eval_step({'obs': np.random.random_sample((2,)), 'legal_actions': {0: None, 1: None}, 'raw_legal_actions': ['call', 'raise']})
self.assertGreaterEqual(predicted_action, 0)
self.assertLessEqual(predicted_action, 1)
for _ in range(num_steps):
ts = [{'obs': np.random.random_sample((2,)), 'legal_actions': {0: None, 1: None}}, np.random.randint(2), 0, {'obs': np.random.random_sample((2,)), 'legal_actions': {0: None, 1: None}, 'raw_legal_actions': ['call', 'raise']}, True]
agent.feed(ts)
predicted_action = agent.step({'obs': np.random.random_sample((2,)), 'legal_actions': {0: None, 1: None}})
self.assertGreaterEqual(predicted_action, 0)
self.assertLessEqual(predicted_action, 1)
def __init__(self,
sess,
scope,
action_num=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=int(1e6),
anticipatory_param=0.1,
batch_size=256,
train_every=1,
rl_learning_rate=0.1,
sl_learning_rate=0.005,
min_buffer_size_to_learn=1000,
q_replay_memory_size=30000,
q_replay_memory_init_size=1000,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=0.06,
q_epsilon_end=0,
q_epsilon_decay_steps=int(1e6),
q_batch_size=256,
q_train_every=1,
q_mlp_layers=None,
evaluate_with='average_policy'):
''' Initialize the NFSP agent.
Args:
sess (tf.Session): Tensorflow session object.
scope (string): The name scope of NFSPAgent.
action_num (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
train_every (int): Train the SL policy every X steps.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_train_step (int): Train the model every X steps.
q_mlp_layers (list): The layer sizes of inner DQN agent.
evaluate_with (string): The value can be 'best_response' or 'average_policy'
'''
self.use_raw = False
self._sess = sess
self._scope = scope
self._action_num = action_num
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes
self._batch_size = batch_size
self._train_every = train_every
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
self.evaluate_with = evaluate_with
self.d = {
0: 'A',
1: '2',
2: '3',
3: '4',
4: '5',
5: '6',
6: '7',
7: '8',
8: '9',
9: 'T',
10: 'J',
11: 'Q',
12: 'K'
}
self.s = {0: 's', 1: 'h', 2: 'd', 3: 'c'}
self.c2n = {
'2': 2,
'3': 3,
'4': 4,
'5': 5,
'6': 6,
'7': 7,
'8': 8,
'9': 9,
'T': 10,
'J': 11,
'Q': 12,
'K': 13,
'A': 14
}
self.late_range = Range(
'22+, A2s+, K2s+, Q2s+, J2s+, J8, T9, 98, 87, 76s, 65s, 54s, 98s+, K9+, Q8+, J7+, T6s+, A9+'
)
# Total timesteps
self.total_t = 0
# Step counter to keep track of learning.
self._step_counter = 0
with tf.variable_scope(scope):
# Inner RL agent
self._rl_agent = DQNAgent(
sess, scope + '_dqn', q_replay_memory_size,
q_replay_memory_init_size, q_update_target_estimator_every,
q_discount_factor, q_epsilon_start, q_epsilon_end,
q_epsilon_decay_steps, q_batch_size, action_num, state_shape,
q_train_every, q_mlp_layers, rl_learning_rate)
with tf.variable_scope('sl'):
# Build supervised model
self._build_model()
self.sample_episode_policy()
class NFSPAgent(object):
''' NFSP Agent implementation in TensorFlow.
'''
def __init__(self,
sess,
scope,
action_num=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=int(1e6),
anticipatory_param=0.1,
batch_size=256,
train_every=1,
rl_learning_rate=0.1,
sl_learning_rate=0.005,
min_buffer_size_to_learn=1000,
q_replay_memory_size=30000,
q_replay_memory_init_size=1000,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=0.06,
q_epsilon_end=0,
q_epsilon_decay_steps=int(1e6),
q_batch_size=256,
q_train_every=1,
q_mlp_layers=None,
evaluate_with='average_policy'):
''' Initialize the NFSP agent.
Args:
sess (tf.Session): Tensorflow session object.
scope (string): The name scope of NFSPAgent.
action_num (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
train_every (int): Train the SL policy every X steps.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_train_step (int): Train the model every X steps.
q_mlp_layers (list): The layer sizes of inner DQN agent.
evaluate_with (string): The value can be 'best_response' or 'average_policy'
'''
self.use_raw = False
self._sess = sess
self._scope = scope
self._action_num = action_num
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes
self._batch_size = batch_size
self._train_every = train_every
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
self.evaluate_with = evaluate_with
self.d = {
0: 'A',
1: '2',
2: '3',
3: '4',
4: '5',
5: '6',
6: '7',
7: '8',
8: '9',
9: 'T',
10: 'J',
11: 'Q',
12: 'K'
}
self.s = {0: 's', 1: 'h', 2: 'd', 3: 'c'}
self.c2n = {
'2': 2,
'3': 3,
'4': 4,
'5': 5,
'6': 6,
'7': 7,
'8': 8,
'9': 9,
'T': 10,
'J': 11,
'Q': 12,
'K': 13,
'A': 14
}
self.late_range = Range(
'22+, A2s+, K2s+, Q2s+, J2s+, J8, T9, 98, 87, 76s, 65s, 54s, 98s+, K9+, Q8+, J7+, T6s+, A9+'
)
# Total timesteps
self.total_t = 0
# Step counter to keep track of learning.
self._step_counter = 0
with tf.variable_scope(scope):
# Inner RL agent
self._rl_agent = DQNAgent(
sess, scope + '_dqn', q_replay_memory_size,
q_replay_memory_init_size, q_update_target_estimator_every,
q_discount_factor, q_epsilon_start, q_epsilon_end,
q_epsilon_decay_steps, q_batch_size, action_num, state_shape,
q_train_every, q_mlp_layers, rl_learning_rate)
with tf.variable_scope('sl'):
# Build supervised model
self._build_model()
self.sample_episode_policy()
def _build_model(self):
''' build the model for supervised learning
'''
# Placeholders.
input_shape = [None]
input_shape.extend(self._state_shape)
self._info_state_ph = tf.placeholder(shape=input_shape,
dtype=tf.float32)
self._X = tf.contrib.layers.flatten(self._info_state_ph)
# Boolean to indicate whether is training or not
self.is_train = tf.placeholder(tf.bool, name="is_train")
# Batch Normalization
self._X = tf.layers.batch_normalization(self._X, training=True)
self._action_probs_ph = tf.placeholder(shape=[None, self._action_num],
dtype=tf.float32)
# Average policy network.
fc = self._X
for dim in self._layer_sizes:
fc = tf.contrib.layers.fully_connected(fc,
dim,
activation_fn=tf.tanh)
self._avg_policy = tf.contrib.layers.fully_connected(
fc, self._action_num, activation_fn=None)
self._avg_policy_probs = tf.nn.softmax(self._avg_policy)
# Loss
self._loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(
labels=tf.stop_gradient(self._action_probs_ph),
logits=self._avg_policy))
optimizer = tf.train.AdamOptimizer(
learning_rate=self._sl_learning_rate, name='nfsp_adam')
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS,
scope=tf.get_variable_scope().name)
with tf.control_dependencies(update_ops):
self._learn_step = optimizer.minimize(self._loss)
def feed(self, ts):
''' Feed data to inner RL agent
Args:
ts (list): A list of 5 elements that represent the transition.
'''
self._rl_agent.feed(ts)
self.total_t += 1
if self.total_t > 0 and len(
self._reservoir_buffer
) >= self._min_buffer_size_to_learn and self.total_t % self._train_every == 0:
sl_loss = self.train_sl()
print('\rINFO - Agent {}, step {}, sl-loss: {}'.format(
self._scope, self.total_t, sl_loss),
end='')
def step(self, state):
''' Returns the action to be taken.
Args:
state (dict): The current state
Returns:
action (int): An action id
'''
self.sample_episode_policy()
cards = ''
pos = 0
for i in state['obs']:
if (i == 1 and pos < 52):
cards = cards + self.d[pos % 13] + '' + self.s[pos // 13]
pos += 1
# if(len(cards) == 4 and not Combo(cards) in self.late_range.combos):
# return 0, 1
tab = []
handcards = cards
for i in state['public_cards']:
tab.append((self.c2n[i[1]], i[0].lower()))
hand = []
for i in range(0, len(handcards), 2):
hand.append((self.c2n[handcards[i]], handcards[i + 1]))
# print(tab)
hand = [x for x in hand if x not in tab]
stt = mcst.PokerState(hand, tab, state['cur'], state['opp'],
abs(state['obs'][-2] - state['obs'][-1]),
state['obs'][-2] + state['obs'][-1],
state['obs'][52], state['obs'][53])
# print(hand, tab, 250 - min(state['obs'][-2:]), 250 - max(state['obs'][-2:]), abs(state['obs'][-2] - state['obs'][-1]), state['obs'][-2] + state['obs'][-1], min(state['obs'][-2:]), max(state['obs'][-2:]))
# mcst.PokerState()
obs = state['obs']
legal_actions = state['legal_actions']
par = mcst.MCTS(1)
if self._mode == MODE.best_response:
probs = self._rl_agent.predict(obs)
m = par.UCT(rootstate=stt,
itermax=50000,
processes=16,
verbose=False)
m = m[0]
probs[m] += 1
probs = remove_illegal(probs, legal_actions)
probs /= sum(probs)
elif self._mode == MODE.average_policy:
probs = self._act(obs)
one_hot = np.eye(len(probs))[np.argmax(probs)]
self._add_transition(obs, one_hot)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
# print(m, action)
return action
def eval_step(self, state):
''' Use the average policy for evaluation purpose
Args:
state (dict): The current state.
Returns:
action (int): An action id.
probs (list): The list of action probabilies
'''
cards = ''
pos = 0
for i in state['obs']:
if (i == 1 and pos < 52):
cards = cards + self.d[pos % 13] + '' + self.s[pos // 13]
pos += 1
# if(len(cards) == 4 and not Combo(cards) in self.late_range.combos):
# return 0, 1
tab = []
handcards = cards
legal_actions = state['legal_actions']
for i in state['public_cards']:
tab.append((self.c2n[i[1]], i[0].lower()))
hand = []
for i in range(0, len(handcards), 2):
hand.append((self.c2n[handcards[i]], handcards[i + 1]))
hand = [x for x in hand if x not in tab]
stt = mcst.PokerState(hand, tab, state['cur'], state['opp'],
abs(state['obs'][-2] - state['obs'][-1]),
state['obs'][-2] + state['obs'][-1],
state['obs'][52], state['obs'][53])
par = mcst.MCTS(1)
# print(state)
if self.evaluate_with == 'best_response':
action, probs = self._rl_agent.eval_step(state)
m = par.UCT(rootstate=stt,
itermax=100000,
processes=32,
verbose=False)
print(m, probs)
m = m[0]
probs[m] += 1
# if probs[1] == probs[3] and probs[3] == probs[4] and probs[4] == probs[5]:
# probs[2] /= 25
# probs[m] += 2
# elif not m == 5:
# probs[m] += 2
# else:
# probs[4] += 3
# if(len(tab) == 0):
# probs[5] = 0
# else:
# probs[5] /= 4
probs = remove_illegal(probs, legal_actions)
probs /= sum(probs)
elif self.evaluate_with == 'average_policy':
obs = state['obs']
probs = self._act(obs)
else:
raise ValueError(
"'evaluate_with' should be either 'average_policy' or 'best_response'."
)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
if (action == 0 and 1 in legal_actions):
action = 1
# print(action, probs)
return action, probs
def sample_episode_policy(self):
''' Sample average/best_response policy
'''
if np.random.rand() < self._anticipatory_param:
self._mode = MODE.best_response
else:
self._mode = MODE.average_policy
def _act(self, info_state):
''' Predict action probability givin the observation and legal actions
Args:
info_state (numpy.array): An obervation.
Returns:
action_probs (numpy.array): The predicted action probability.
'''
info_state = np.expand_dims(info_state, axis=0)
action_probs = self._sess.run(self._avg_policy_probs,
feed_dict={
self._info_state_ph: info_state,
self.is_train: False
})[0]
return action_probs
def _add_transition(self, state, probs):
''' Adds the new transition to the reservoir buffer.
Transitions are in the form (state, probs).
Args:
state (numpy.array): The state.
probs (numpy.array): The probabilities of each action.
'''
transition = Transition(info_state=state, action_probs=probs)
self._reservoir_buffer.add(transition)
def train_sl(self):
''' Compute the loss on sampled transitions and perform a avg-network update.
If there are not enough elements in the buffer, no loss is computed and
`None` is returned instead.
Returns:
loss (float): The average loss obtained on this batch of transitions or `None`.
'''
if (len(self._reservoir_buffer) < self._batch_size or
len(self._reservoir_buffer) < self._min_buffer_size_to_learn):
return None
transitions = self._reservoir_buffer.sample(self._batch_size)
info_states = [t.info_state for t in transitions]
action_probs = [t.action_probs for t in transitions]
loss, _ = self._sess.run(
[self._loss, self._learn_step],
feed_dict={
self._info_state_ph: info_states,
self._action_probs_ph: action_probs,
self.is_train: True,
})
return loss
memory_init_size = 1000
norm_step = 100
# The paths for saving the logs and learning curves
root_path = './experiments/uno_dqnvsrandom_result/'
log_path = root_path + 'log.txt'
csv_path = root_path + 'performance.csv'
figure_path = root_path + 'figures/'
## Set a global seed
##set_global_seed(0)
if True:
sess1 = tf.compat.v1.Session()
# Set agents
global_step = tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(sess1,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[100, 100])
sess1.run(tf.global_variables_initializer())
saver = tf.train.Saver()
saver.restore(sess1, root_path + 'models/model1.ckpt')
saver.save(sess1, './experiments/uno_dqn_result/models/modelvrandom1.ckpt')
env.set_agents([agent])
ishuman = [-1, 0]
env.gamewithai(ishuman)
# The paths for saving the logs and learning curves
root_path = './experiments/leduc_holdem_single_agent_dqn_result/'
log_path = root_path + 'log.txt'
csv_path = root_path + 'performance.csv'
figure_path = root_path + 'figures/'
# Set a global seed
set_global_seed(0)
with tf.Session() as sess:
global_step = tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_size=int(1e5),
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[128, 128])
sess.run(tf.global_variables_initializer())
# Init a Logger to plot the learning curve
logger = Logger(xlabel='timestep',
ylabel='reward',
legend='DQN on Leduc Holdem',
log_path=log_path,
csv_path=csv_path)
state = env.reset()
def __init__(self,
num_actions=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=20000,
anticipatory_param=0.1,
batch_size=256,
train_every=1,
rl_learning_rate=0.1,
sl_learning_rate=0.005,
min_buffer_size_to_learn=100,
q_replay_memory_size=20000,
q_replay_memory_init_size=100,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=0.06,
q_epsilon_end=0,
q_epsilon_decay_steps=int(1e6),
q_batch_size=32,
q_train_every=1,
q_mlp_layers=None,
evaluate_with='average_policy',
device=None):
''' Initialize the NFSP agent.
Args:
num_actions (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
train_every (int): Train the SL policy every X steps.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_train_step (int): Train the model every X steps.
q_mlp_layers (list): The layer sizes of inner DQN agent.
device (torch.device): Whether to use the cpu or gpu
'''
self.use_raw = False
self._num_actions = num_actions
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes + [num_actions]
self._batch_size = batch_size
self._train_every = train_every
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
self.evaluate_with = evaluate_with
if device is None:
self.device = torch.device(
'cuda:0' if torch.cuda.is_available() else 'cpu')
else:
self.device = device
# Total timesteps
self.total_t = 0
# Step counter to keep track of learning.
self._step_counter = 0
# Build the action-value network
self._rl_agent = DQNAgent(q_replay_memory_size, q_replay_memory_init_size, \
q_update_target_estimator_every, q_discount_factor, q_epsilon_start, q_epsilon_end, \
q_epsilon_decay_steps, q_batch_size, num_actions, state_shape, q_train_every, q_mlp_layers, \
rl_learning_rate, device)
# Build the average policy supervised model
self._build_model()
self.sample_episode_policy()
# The paths for saving the logs and learning curves
log_dir = './experiments/blackjack_dqn_result/'
# Set a global seed
set_global_seed(0)
with tf.Session() as sess:
# Initialize a global step
global_step = tf.Variable(0, name='global_step', trainable=False)
# Set up the agents
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
train_every=train_every,
state_shape=env.state_shape,
mlp_layers=[10,10])
env.set_agents([agent])
eval_env.set_agents([agent])
# Initialize global variables
sess.run(tf.global_variables_initializer())
# Initialize a Logger to plot the learning curve
logger = Logger(log_dir)
for episode in range(episode_num):
# Generate data from the environment
root_path = './experiments/uno_dqn_result/'
log_path = root_path + 'log.txt'
csv_path = root_path + 'performance.csv'
figure_path = root_path + 'figures/'
# Set a global seed
set_global_seed(0)
with tf.Session() as sess:
# Set agents
global_step = tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[10, 400])
env.set_agents([agent, RandomAgent(action_num=env.action_num)])
eval_env.set_agents([agent, RandomAgent(action_num=env.action_num)])
sess.run(tf.global_variables_initializer())
# Count the number of steps
step_counter = 0
# Init a Logger to plot the learning curve
logger = Logger(xlabel='timestep',
ylabel='reward',
legend='DQN on UNO',
# The paths for saving the logs and learning curves
log_dir = './experiments/uno_dqn_result/'
# Set a global seed
set_global_seed(0)
with tf.Session() as sess:
# Initialize a global step
global_step = tf.Variable(0, name='global_step', trainable=False)
# Set up the agents
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_size=20000,
replay_memory_init_size=memory_init_size,
state_shape=env.state_shape,
mlp_layers=[512, 512])
random_agent = RandomAgent(action_num=eval_env.action_num)
env.set_agents([agent, random_agent, random_agent, random_agent])
eval_env.set_agents([agent, random_agent, random_agent, random_agent])
# Initialize global variables
sess.run(tf.global_variables_initializer())
# Init a Logger to plot the learning curve
logger = Logger(log_dir)
for episode in range(episode_num):
root_path = './experiments/badugi_hand_dqn_result/'
log_path = root_path + 'log.txt'
csv_path = root_path + 'performance.csv'
figure_path = root_path + 'figures/'
checkpoint_path = root_path + 'ckpt/'
# Set a global seed
# set_global_seed(0)
with tf.Session() as sess:
# Set agents
global_step = tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_size=int(1e5),
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[512, 512])
random_agent = RandomAgent(action_num=eval_env.action_num)
sess.run(tf.global_variables_initializer())
env.set_agents([agent, random_agent, random_agent, random_agent, random_agent])
eval_env.set_agents([agent, random_agent, random_agent, random_agent, random_agent])
# Count the number of steps
step_counter = 0
# Init a Logger to plot the learning curve
class NFSPAgent(object):
''' NFSP Agent implementation in TensorFlow.
'''
def __init__(self,
sess,
scope,
action_num=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=int(1e6),
anticipatory_param=0.5,
batch_size=256,
rl_learning_rate=0.0001,
sl_learning_rate=0.00001,
min_buffer_size_to_learn=1000,
q_replay_memory_size=30000,
q_replay_memory_init_size=1000,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=1,
q_epsilon_end=0.1,
q_epsilon_decay_steps=int(1e6),
q_batch_size=256,
q_norm_step=1000,
q_mlp_layers=None):
''' Initialize the NFSP agent.
Args:
sess (tf.Session): Tensorflow session object.
scope (string): The name scope of NFSPAgent.
action_num (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_norm_step (int): The normalization steps of inner DQN agent.
q_mlp_layers (list): The layer sizes of inner DQN agent.
'''
self._sess = sess
self._action_num = action_num
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes + [action_num]
self._batch_size = batch_size
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
# Step counter to keep track of learning.
self._step_counter = 0
with tf.variable_scope(scope):
# Inner RL agent
self._rl_agent = DQNAgent(
sess, 'dqn', q_replay_memory_size, q_replay_memory_init_size,
q_update_target_estimator_every, q_discount_factor,
q_epsilon_start, q_epsilon_end, q_epsilon_decay_steps,
q_batch_size, action_num, state_shape, q_norm_step,
q_mlp_layers, rl_learning_rate)
# Build supervised model
self._build_model()
self.sample_episode_policy()
def _build_model(self):
''' build the model for supervised learning
'''
# Placeholders.
input_shape = [None]
input_shape.extend(self._state_shape)
self._info_state_ph = tf.placeholder(shape=input_shape,
dtype=tf.float32)
self._X = tf.contrib.layers.flatten(self._info_state_ph)
self._action_probs_ph = tf.placeholder(shape=[None, self._action_num],
dtype=tf.float32)
# Average policy network.
self._avg_network = snt.nets.MLP(output_sizes=self._layer_sizes)
self._avg_policy = self._avg_network(self._X)
self._avg_policy_probs = tf.nn.softmax(self._avg_policy)
# Loss
self._loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(
labels=tf.stop_gradient(self._action_probs_ph),
logits=self._avg_policy))
optimizer = tf.train.AdamOptimizer(
learning_rate=self._sl_learning_rate, name='nfsp_adam')
self._learn_step = optimizer.minimize(self._loss)
def feed(self, ts):
''' Feed data to inner RL agent
Args:
ts (list): A list of 5 elements that represent the transition.
'''
self._rl_agent.feed(ts)
def step(self, state):
''' Returns the action to be taken.
Args:
state (dict): The current state
Returns:
action (int): An action id
'''
obs = state['obs']
legal_actions = state['legal_actions']
if self._mode == MODE.best_response:
probs = self._rl_agent.predict(obs)
self._add_transition(obs, probs)
elif self._mode == MODE.average_policy:
probs = self._act(obs)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
return action
def eval_step(self, state):
''' Use the average policy for evaluation purpose
Args:
state (dict): The current state.
Returns:
action (int): An action id.
'''
action = self._rl_agent.eval_step(state)
return action
def sample_episode_policy(self):
''' Sample average/best_response policy
'''
if np.random.rand() < self._anticipatory_param:
self._mode = MODE.best_response
else:
self._mode = MODE.average_policy
def _act(self, info_state):
''' Predict action probability givin the observation and legal actions
Args:
info_state (numpy.array): An obervation.
Returns:
action_probs (numpy.array): The predicted action probability.
'''
info_state = np.expand_dims(info_state, axis=0)
action_probs = self._sess.run(
self._avg_policy_probs,
feed_dict={self._info_state_ph: info_state})[0]
return action_probs
def _add_transition(self, state, probs):
''' Adds the new transition to the reservoir buffer.
Transitions are in the form (state, probs).
Args:
state (numpy.array): The state.
probs (numpy.array): The probabilities of each action.
'''
#print(len(self._reservoir_buffer))
transition = Transition(info_state=state, action_probs=probs)
self._reservoir_buffer.add(transition)
def train_rl(self):
''' Update the inner RL agent
'''
return self._rl_agent.train()
def train_sl(self):
''' Compute the loss on sampled transitions and perform a avg-network update.
If there are not enough elements in the buffer, no loss is computed and
`None` is returned instead.
Returns:
loss (float): The average loss obtained on this batch of transitions or `None`.
'''
if (len(self._reservoir_buffer) < self._batch_size or
len(self._reservoir_buffer) < self._min_buffer_size_to_learn):
return None
transitions = self._reservoir_buffer.sample(self._batch_size)
info_states = [t.info_state for t in transitions]
action_probs = [t.action_probs for t in transitions]
loss, _ = self._sess.run(
[self._loss, self._learn_step],
feed_dict={
self._info_state_ph: info_states,
self._action_probs_ph: action_probs,
})
return loss
model_path = 'rlcard/models/pretrained/self_played_{}/tarot_v{}/model'.format(
str(record_number), str(record_number * 10000 + self_play))
# Set a global seed
set_global_seed(0)
random_agent = RandomAgent(action_num=eval_env.action_num)
with tf.compat.v1.Session() as sess:
# Set agents
global_step = tf.Variable(0, name='global_step', trainable=False)
agent = DQNAgent(
sess,
scope='dqn',
action_num=78, # env.action_num,
replay_memory_size=20000,
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=env.state_shape,
mlp_layers=[512, 1024, 512])
opponent_agent = agent
sess.run(tf.compat.v1.global_variables_initializer())
saver = tf.compat.v1.train.Saver()
env.set_agents([agent] + [opponent_agent] * (env.player_num - 1))
eval_env.set_agents([agent] + [random_agent] * (env.player_num - 1))
# Count the number of steps
evaluate_num = 1000
episode_num = 1000000
# Set the the number of steps for collecting normalization statistics
# and intial memory size
memory_init_size = 100
norm_step = 100
# Set a global seed
set_global_seed(1)
with tf.Session() as sess:
# Set agents
agent = DQNAgent(sess,
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
norm_step=norm_step,
state_shape=[2],
mlp_layers=[10,10])
env.set_agents([agent])
# Count the number of steps
step_counter = 0
# Init a Logger to plot the learning curve
logger = Logger(xlabel='eposide', ylabel='reward', legend='DQN on Blackjack', log_path='./experiments/blackjack_dqn_result/log.txt', csv_path='./experiments/blackjack_dqn_result/performance.csv')
for episode in range(episode_num):
# Generate data from the environment
trajectories, _ = env.run(is_training=True)
def __init__(self,
sess,
scope,
action_num=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=int(1e6),
anticipatory_param=0.1,
batch_size=256,
train_every=1,
rl_learning_rate=0.1,
sl_learning_rate=0.005,
min_buffer_size_to_learn=1000,
q_replay_memory_size=30000,
q_replay_memory_init_size=1000,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=0.06,
q_epsilon_end=0,
q_epsilon_decay_steps=int(1e6),
q_batch_size=256,
q_train_every=1,
q_mlp_layers=None,
evaluate_with='average_policy'):
''' Initialize the NFSP agent.
Args:
sess (tf.Session): Tensorflow session object.
scope (string): The name scope of NFSPAgent.
action_num (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
train_every (int): Train the SL policy every X steps.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_train_step (int): Train the model every X steps.
q_mlp_layers (list): The layer sizes of inner DQN agent.
evaluate_with (string): The value can be 'best_response' or 'average_policy'
'''
self.use_raw = False
self._sess = sess
self._scope = scope
self._action_num = action_num
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes
self._batch_size = batch_size
self._train_every = train_every
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
self.evaluate_with = evaluate_with
# Total timesteps
self.total_t = 0
# Step counter to keep track of learning.
self._step_counter = 0
with tf.compat.v1.variable_scope(scope):
# Inner RL agent
self._rl_agent = DQNAgent(
sess, scope + '_dqn', q_replay_memory_size,
q_replay_memory_init_size, q_update_target_estimator_every,
q_discount_factor, q_epsilon_start, q_epsilon_end,
q_epsilon_decay_steps, q_batch_size, action_num, state_shape,
q_train_every, q_mlp_layers, rl_learning_rate)
with tf.compat.v1.variable_scope('sl'):
# Build supervised model
self._build_model()
self.sample_episode_policy()
class NFSPAgent(object):
''' An approximate clone of rlcard.agents.nfsp_agent that uses
pytorch instead of tensorflow. Note that this implementation
differs from Henrich and Silver (2016) in that the supervised
training minimizes cross-entropy with respect to the stored
action probabilities rather than the realized actions.
'''
def __init__(self,
num_actions=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=20000,
anticipatory_param=0.1,
batch_size=256,
train_every=1,
rl_learning_rate=0.1,
sl_learning_rate=0.005,
min_buffer_size_to_learn=100,
q_replay_memory_size=20000,
q_replay_memory_init_size=100,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=0.06,
q_epsilon_end=0,
q_epsilon_decay_steps=int(1e6),
q_batch_size=32,
q_train_every=1,
q_mlp_layers=None,
evaluate_with='average_policy',
device=None):
''' Initialize the NFSP agent.
Args:
num_actions (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
train_every (int): Train the SL policy every X steps.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_train_step (int): Train the model every X steps.
q_mlp_layers (list): The layer sizes of inner DQN agent.
device (torch.device): Whether to use the cpu or gpu
'''
self.use_raw = False
self._num_actions = num_actions
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes + [num_actions]
self._batch_size = batch_size
self._train_every = train_every
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
self.evaluate_with = evaluate_with
if device is None:
self.device = torch.device(
'cuda:0' if torch.cuda.is_available() else 'cpu')
else:
self.device = device
# Total timesteps
self.total_t = 0
# Step counter to keep track of learning.
self._step_counter = 0
# Build the action-value network
self._rl_agent = DQNAgent(q_replay_memory_size, q_replay_memory_init_size, \
q_update_target_estimator_every, q_discount_factor, q_epsilon_start, q_epsilon_end, \
q_epsilon_decay_steps, q_batch_size, num_actions, state_shape, q_train_every, q_mlp_layers, \
rl_learning_rate, device)
# Build the average policy supervised model
self._build_model()
self.sample_episode_policy()
def _build_model(self):
''' Build the average policy network
'''
# configure the average policy network
policy_network = AveragePolicyNetwork(self._num_actions,
self._state_shape,
self._layer_sizes)
policy_network = policy_network.to(self.device)
self.policy_network = policy_network
self.policy_network.eval()
# xavier init
for p in self.policy_network.parameters():
if len(p.data.shape) > 1:
nn.init.xavier_uniform_(p.data)
# configure optimizer
self.policy_network_optimizer = torch.optim.Adam(
self.policy_network.parameters(), lr=self._sl_learning_rate)
def feed(self, ts):
''' Feed data to inner RL agent
Args:
ts (list): A list of 5 elements that represent the transition.
'''
self._rl_agent.feed(ts)
self.total_t += 1
if self.total_t > 0 and len(
self._reservoir_buffer
) >= self._min_buffer_size_to_learn and self.total_t % self._train_every == 0:
sl_loss = self.train_sl()
print('\rINFO - Step {}, sl-loss: {}'.format(
self.total_t, sl_loss),
end='')
def step(self, state):
''' Returns the action to be taken.
Args:
state (dict): The current state
Returns:
action (int): An action id
'''
obs = state['obs']
legal_actions = list(state['legal_actions'].keys())
if self._mode == 'best_response':
action = self._rl_agent.step(state)
one_hot = np.zeros(self._num_actions)
one_hot[action] = 1
self._add_transition(obs, one_hot)
elif self._mode == 'average_policy':
probs = self._act(obs)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
return action
def eval_step(self, state):
''' Use the average policy for evaluation purpose
Args:
state (dict): The current state.
Returns:
action (int): An action id.
info (dict): A dictionary containing information
'''
if self.evaluate_with == 'best_response':
action, info = self._rl_agent.eval_step(state)
elif self.evaluate_with == 'average_policy':
obs = state['obs']
legal_actions = list(state['legal_actions'].keys())
probs = self._act(obs)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
info = {}
info['probs'] = {
state['raw_legal_actions'][i]:
float(probs[list(state['legal_actions'].keys())[i]])
for i in range(len(state['legal_actions']))
}
else:
raise ValueError(
"'evaluate_with' should be either 'average_policy' or 'best_response'."
)
return action, info
def sample_episode_policy(self):
''' Sample average/best_response policy
'''
if np.random.rand() < self._anticipatory_param:
self._mode = 'best_response'
else:
self._mode = 'average_policy'
def _act(self, info_state):
''' Predict action probability givin the observation and legal actions
Not connected to computation graph
Args:
info_state (numpy.array): An obervation.
Returns:
action_probs (numpy.array): The predicted action probability.
'''
info_state = np.expand_dims(info_state, axis=0)
info_state = torch.from_numpy(info_state).float().to(self.device)
with torch.no_grad():
log_action_probs = self.policy_network(info_state).cpu().numpy()
action_probs = np.exp(log_action_probs)[0]
return action_probs
def _add_transition(self, state, probs):
''' Adds the new transition to the reservoir buffer.
Transitions are in the form (state, probs).
Args:
state (numpy.array): The state.
probs (numpy.array): The probabilities of each action.
'''
transition = Transition(info_state=state, action_probs=probs)
self._reservoir_buffer.add(transition)
def train_sl(self):
''' Compute the loss on sampled transitions and perform a avg-network update.
If there are not enough elements in the buffer, no loss is computed and
`None` is returned instead.
Returns:
loss (float): The average loss obtained on this batch of transitions or `None`.
'''
if (len(self._reservoir_buffer) < self._batch_size or
len(self._reservoir_buffer) < self._min_buffer_size_to_learn):
return None
transitions = self._reservoir_buffer.sample(self._batch_size)
info_states = [t.info_state for t in transitions]
action_probs = [t.action_probs for t in transitions]
self.policy_network_optimizer.zero_grad()
self.policy_network.train()
# (batch, state_size)
info_states = torch.from_numpy(np.array(info_states)).float().to(
self.device)
# (batch, num_actions)
eval_action_probs = torch.from_numpy(
np.array(action_probs)).float().to(self.device)
# (batch, num_actions)
log_forecast_action_probs = self.policy_network(info_states)
ce_loss = -(eval_action_probs *
log_forecast_action_probs).sum(dim=-1).mean()
ce_loss.backward()
self.policy_network_optimizer.step()
ce_loss = ce_loss.item()
self.policy_network.eval()
return ce_loss
def set_device(self, device):
self.device = device
self._rl_agent.set_device(device)
# The paths for saving the logs and learning curves
log_dir = './experiments/uno_single_dqn_result/'
# Set a global seed
set_global_seed(0)
with tf.Session() as sess:
# Initialize a global step
global_step = tf.Variable(0, name='global_step', trainable=False)
# Set up the agents
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
train_every=train_every,
state_shape=env.state_shape,
mlp_layers=[128, 128])
# Initialize global variables
sess.run(tf.global_variables_initializer())
# Init a Logger to plot the learning curve
logger = Logger(log_dir)
state = env.reset()
for timestep in range(timesteps):
action = agent.step(state)
next_state, reward, done = env.step(action)
ts = (state, action, reward, next_state, done)
class NFSPAgent(object):
''' NFSP Agent implementation in TensorFlow.
'''
def __init__(self,
sess,
scope,
action_num=4,
state_shape=None,
hidden_layers_sizes=None,
reservoir_buffer_capacity=int(1e6),
anticipatory_param=0.1,
batch_size=256,
train_every=1,
rl_learning_rate=0.1,
sl_learning_rate=0.005,
min_buffer_size_to_learn=1000,
q_replay_memory_size=30000,
q_replay_memory_init_size=1000,
q_update_target_estimator_every=1000,
q_discount_factor=0.99,
q_epsilon_start=0.06,
q_epsilon_end=0,
q_epsilon_decay_steps=int(1e6),
q_batch_size=256,
q_train_every=1,
q_mlp_layers=None,
evaluate_with='average_policy'):
''' Initialize the NFSP agent.
Args:
sess (tf.Session): Tensorflow session object.
scope (string): The name scope of NFSPAgent.
action_num (int): The number of actions.
state_shape (list): The shape of the state space.
hidden_layers_sizes (list): The hidden layers sizes for the layers of
the average policy.
reservoir_buffer_capacity (int): The size of the buffer for average policy.
anticipatory_param (float): The hyper-parameter that balances rl/avarage policy.
batch_size (int): The batch_size for training average policy.
train_every (int): Train the SL policy every X steps.
rl_learning_rate (float): The learning rate of the RL agent.
sl_learning_rate (float): the learning rate of the average policy.
min_buffer_size_to_learn (int): The minimum buffer size to learn for average policy.
q_replay_memory_size (int): The memory size of inner DQN agent.
q_replay_memory_init_size (int): The initial memory size of inner DQN agent.
q_update_target_estimator_every (int): The frequency of updating target network for
inner DQN agent.
q_discount_factor (float): The discount factor of inner DQN agent.
q_epsilon_start (float): The starting epsilon of inner DQN agent.
q_epsilon_end (float): the end epsilon of inner DQN agent.
q_epsilon_decay_steps (int): The decay steps of inner DQN agent.
q_batch_size (int): The batch size of inner DQN agent.
q_train_step (int): Train the model every X steps.
q_mlp_layers (list): The layer sizes of inner DQN agent.
evaluate_with (string): The value can be 'best_response' or 'average_policy'
'''
self.use_raw = False
self._sess = sess
self._scope = scope
self._action_num = action_num
self._state_shape = state_shape
self._layer_sizes = hidden_layers_sizes
self._batch_size = batch_size
self._train_every = train_every
self._sl_learning_rate = sl_learning_rate
self._anticipatory_param = anticipatory_param
self._min_buffer_size_to_learn = min_buffer_size_to_learn
self._reservoir_buffer = ReservoirBuffer(reservoir_buffer_capacity)
self._prev_timestep = None
self._prev_action = None
self.evaluate_with = evaluate_with
# Total timesteps
self.total_t = 0
# Step counter to keep track of learning.
self._step_counter = 0
with tf.compat.v1.variable_scope(scope):
# Inner RL agent
self._rl_agent = DQNAgent(
sess, scope + '_dqn', q_replay_memory_size,
q_replay_memory_init_size, q_update_target_estimator_every,
q_discount_factor, q_epsilon_start, q_epsilon_end,
q_epsilon_decay_steps, q_batch_size, action_num, state_shape,
q_train_every, q_mlp_layers, rl_learning_rate)
with tf.compat.v1.variable_scope('sl'):
# Build supervised model
self._build_model()
self.sample_episode_policy()
def _build_model(self):
''' build the model for supervised learning
'''
# Placeholders.
input_shape = [None]
input_shape.extend(self._state_shape)
self._info_state_ph = tf.compat.v1.placeholder(shape=input_shape,
dtype=tf.float32)
self._X = tf.keras.layers.Flatten(self._info_state_ph)
# Boolean to indicate whether is training or not
self.is_train = tf.compat.v1.placeholder(tf.bool, name="is_train")
# Batch Normalization
self._X = tf.compat.v1.layers.batch_normalization(self._X,
training=True)
self._action_probs_ph = tf.compat.v1.placeholder(
shape=[None, self._action_num], dtype=tf.float32)
# Average policy network.
fc = self._X
# to be fixed
for dim in self._layer_sizes:
fc.add(tf.keras.layers.Dense(dim, activation=tf.tanh))
self._avg_policy = tf.keras.layers.Dense(self._action_num,
activation=None)
self._avg_policy_probs = tf.nn.softmax(self._avg_policy)
# Loss
self._loss = tf.reduce_mean(
input_tensor=tf.nn.softmax_cross_entropy_with_logits(
labels=tf.stop_gradient(self._action_probs_ph),
logits=self._avg_policy))
optimizer = tf.compat.v1.train.AdamOptimizer(
learning_rate=self._sl_learning_rate, name='nfsp_adam')
update_ops = tf.compat.v1.get_collection(
tf.compat.v1.GraphKeys.UPDATE_OPS,
scope=tf.compat.v1.get_variable_scope().name)
with tf.control_dependencies(update_ops):
self._learn_step = optimizer.minimize(self._loss)
def feed(self, ts):
''' Feed data to inner RL agent
Args:
ts (list): A list of 5 elements that represent the transition.
'''
self._rl_agent.feed(ts)
self.total_t += 1
if self.total_t > 0 and len(
self._reservoir_buffer
) >= self._min_buffer_size_to_learn and self.total_t % self._train_every == 0:
sl_loss = self.train_sl()
print('\rINFO - Agent {}, step {}, sl-loss: {}'.format(
self._scope, self.total_t, sl_loss),
end='')
def step(self, state):
''' Returns the action to be taken.
Args:
state (dict): The current state
Returns:
action (int): An action id
'''
obs = state['obs']
legal_actions = state['legal_actions']
if self._mode == MODE.best_response:
probs = self._rl_agent.predict(obs)
one_hot = np.eye(len(probs))[np.argmax(probs)]
self._add_transition(obs, one_hot)
elif self._mode == MODE.average_policy:
probs = self._act(obs)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
return action
def eval_step(self, state):
''' Use the average policy for evaluation purpose
Args:
state (dict): The current state.
Returns:
action (int): An action id.
probs (list): The list of action probabilies
'''
if self.evaluate_with == 'best_response':
action, probs = self._rl_agent.eval_step(state)
elif self.evaluate_with == 'average_policy':
obs = state['obs']
legal_actions = state['legal_actions']
probs = self._act(obs)
probs = remove_illegal(probs, legal_actions)
action = np.random.choice(len(probs), p=probs)
else:
raise ValueError(
"'evaluate_with' should be either 'average_policy' or 'best_response'."
)
return action, probs
def sample_episode_policy(self):
''' Sample average/best_response policy
'''
if np.random.rand() < self._anticipatory_param:
self._mode = MODE.best_response
else:
self._mode = MODE.average_policy
def _act(self, info_state):
''' Predict action probability givin the observation and legal actions
Args:
info_state (numpy.array): An obervation.
Returns:
action_probs (numpy.array): The predicted action probability.
'''
info_state = np.expand_dims(info_state, axis=0)
action_probs = self._sess.run(self._avg_policy_probs,
feed_dict={
self._info_state_ph: info_state,
self.is_train: False
})[0]
return action_probs
def _add_transition(self, state, probs):
''' Adds the new transition to the reservoir buffer.
Transitions are in the form (state, probs).
Args:
state (numpy.array): The state.
probs (numpy.array): The probabilities of each action.
'''
transition = Transition(info_state=state, action_probs=probs)
self._reservoir_buffer.add(transition)
def train_sl(self):
''' Compute the loss on sampled transitions and perform a avg-network update.
If there are not enough elements in the buffer, no loss is computed and
`None` is returned instead.
Returns:
loss (float): The average loss obtained on this batch of transitions or `None`.
'''
if (len(self._reservoir_buffer) < self._batch_size or
len(self._reservoir_buffer) < self._min_buffer_size_to_learn):
return None
transitions = self._reservoir_buffer.sample(self._batch_size)
info_states = [t.info_state for t in transitions]
action_probs = [t.action_probs for t in transitions]
loss, _ = self._sess.run(
[self._loss, self._learn_step],
feed_dict={
self._info_state_ph: info_states,
self._action_probs_ph: action_probs,
self.is_train: True,
})
return loss
