def backward(self, out_grad, in_data, out_data, in_grad):
x = out_data[0]
action = in_data[1] #.astype('int')
reward = in_data[2]
dx = in_grad[0]
dx[:] = 0
dx[:] = nd.fill_element_0index(
dx,
nd.choose_element_0index(x, action) - reward, action)
def backward(self, out_grad, in_data, out_data, in_grad):
x = out_data[0]
action = in_data[1]
reward = in_data[2]
dx = in_grad[0]
dx[:] = 0
dx[:] = nd.fill_element_0index(dx,
nd.clip(nd.choose_element_0index(x, action) - reward, -1, 1),
action)
def train_one_batch(self, st, stpo, at, rt, tt, unfreeze_weight=False):
"""
st : state_at_time_t
stpo : state_at_time_t_plus_one
at : action at time t
rt : reward at time t --> instant reward
tt : termination of the game at time t : 0 or 1
From all the state needed for a forward backward,
each sub result is calculated to reach the loss backward
"""
# 1 : get the y_ddqn
# The formula is Yddqn = Rt + Q(st+1, a*, theta-)
# a) forward on the q vector (net with theta -)
target_q = self.target_q_mod.forward(is_train=False, data=stpo)
# b) retrieve a* with theta_i param : a = argmax Q(stpo, theta_i)
a_q = self.loss_q_mod.forward(is_train=False, data=stpo)
a = nd.argmax_channel(a_q[0])
# c) combine q vec with the a*
y_ddqn = rt + self.gamma * (tt) * nd.choose_element_0index(
target_q[0], a)
# 2 : build the loss on the current batch
# Here some optim is done with the full netwok loss waiting for a y_ddqn
# The loss is just | y_ddqn - Q(st, at) |
current_q = self.loss_q_mod.forward(is_train=True,
data=st,
loss_action=a,
loss_target=y_ddqn)
print(current_q[0], at, rt)
self.loss_q_mod.backward()
# 3 : Update parameters
self.update_weights(self.loss_q_mod, self.updater)
# 4 : Calculate the loss
loss = nd.sum((y_ddqn - nd.choose_element_0index(current_q[0], at))**2,
axis=0)
# 5 : the occasional forward weight updater
if unfreeze_weight:
self.copy_to_freezed_network()
return loss[0]
def main():
parser = argparse.ArgumentParser(
description='Script to test the trained network on a game.')
parser.add_argument('-r',
'--rom',
required=False,
type=str,
default=os.path.join('arena', 'games', 'roms',
'breakout.bin'),
help='Path of the ROM File.')
parser.add_argument('-v',
'--visualization',
required=False,
type=int,
default=0,
help='Visualize the runs.')
parser.add_argument('--lr',
required=False,
type=float,
default=0.01,
help='Learning rate of the AdaGrad optimizer')
parser.add_argument('--eps',
required=False,
type=float,
default=0.01,
help='Eps of the AdaGrad optimizer')
parser.add_argument('--clip-gradient',
required=False,
type=float,
default=None,
help='Clip threshold of the AdaGrad optimizer')
parser.add_argument('--double-q',
required=False,
type=bool,
default=False,
help='Use Double DQN')
parser.add_argument('--wd',
required=False,
type=float,
default=0.0,
help='Weight of the L2 Regularizer')
parser.add_argument(
'-c',
'--ctx',
required=False,
type=str,
default='gpu',
help='Running Context. E.g `-c gpu` or `-c gpu1` or `-c cpu`')
parser.add_argument('-d',
'--dir-path',
required=False,
type=str,
default='',
help='Saving directory of model files.')
parser.add_argument(
'--start-eps',
required=False,
type=float,
default=1.0,
help='Eps of the epsilon-greedy policy at the beginning')
parser.add_argument('--replay-start-size',
required=False,
type=int,
default=50000,
help='The step that the training starts')
parser.add_argument(
'--kvstore-update-period',
required=False,
type=int,
default=1,
help='The period that the worker updates the parameters from the sever'
)
parser.add_argument(
'--kv-type',
required=False,
type=str,
default=None,
help=
'type of kvstore, default will not use kvstore, could also be dist_async'
)
parser.add_argument('--optimizer',
required=False,
type=str,
default="adagrad",
help='type of optimizer')
args = parser.parse_args()
if args.dir_path == '':
rom_name = os.path.splitext(os.path.basename(args.rom))[0]
args.dir_path = 'dqn-%s-lr%g' % (rom_name, args.lr)
replay_start_size = args.replay_start_size
max_start_nullops = 30
replay_memory_size = 1000000
history_length = 4
rows = 84
cols = 84
ctx = parse_ctx(args.ctx)
q_ctx = mx.Context(*ctx[0])
game = AtariGame(rom_path=args.rom,
resize_mode='scale',
replay_start_size=replay_start_size,
resized_rows=rows,
resized_cols=cols,
max_null_op=max_start_nullops,
replay_memory_size=replay_memory_size,
display_screen=args.visualization,
history_length=history_length)
##RUN NATURE
freeze_interval = 10000
epoch_num = 200
steps_per_epoch = 250000
update_interval = 4
discount = 0.99
eps_start = args.start_eps
eps_min = 0.1
eps_decay = (eps_start - eps_min) / 1000000
eps_curr = eps_start
freeze_interval /= update_interval
minibatch_size = 32
action_num = len(game.action_set)
data_shapes = {
'data': (minibatch_size, history_length) + (rows, cols),
'dqn_action': (minibatch_size, ),
'dqn_reward': (minibatch_size, )
}
dqn_sym = dqn_sym_nature(action_num)
qnet = Base(data_shapes=data_shapes,
sym_gen=dqn_sym,
name='QNet',
initializer=DQNInitializer(factor_type="in"),
ctx=q_ctx)
target_qnet = qnet.copy(name="TargetQNet", ctx=q_ctx)
use_easgd = False
if args.optimizer != "easgd":
optimizer = mx.optimizer.create(name=args.optimizer,
learning_rate=args.lr,
eps=args.eps,
clip_gradient=args.clip_gradient,
rescale_grad=1.0,
wd=args.wd)
else:
use_easgd = True
easgd_beta = 0.9
easgd_p = 4
easgd_alpha = easgd_beta / (args.kvstore_update_period * easgd_p)
server_optimizer = mx.optimizer.create(name="ServerEASGD",
learning_rate=easgd_alpha)
easgd_eta = 0.00025
local_optimizer = mx.optimizer.create(name='adagrad',
learning_rate=args.lr,
eps=args.eps,
clip_gradient=args.clip_gradient,
rescale_grad=1.0,
wd=args.wd)
central_weight = OrderedDict([(n, nd.zeros(v.shape, ctx=q_ctx))
for n, v in qnet.params.items()])
# Create KVStore
if args.kv_type != None:
kv = kvstore.create(args.kv_type)
#Initialize KVStore
for idx, v in enumerate(qnet.params.values()):
kv.init(idx, v)
# Set Server optimizer on KVStore
if not use_easgd:
kv.set_optimizer(optimizer)
else:
kv.set_optimizer(server_optimizer)
local_updater = mx.optimizer.get_updater(local_optimizer)
kvstore_update_period = args.kvstore_update_period
args.dir_path = args.dir_path + "-" + str(kv.rank)
else:
updater = mx.optimizer.get_updater(optimizer)
qnet.print_stat()
target_qnet.print_stat()
# Begin Playing Game
training_steps = 0
total_steps = 0
for epoch in xrange(epoch_num):
# Run Epoch
steps_left = steps_per_epoch
episode = 0
epoch_reward = 0
start = time.time()
game.start()
while steps_left > 0:
# Running New Episode
episode += 1
episode_loss = 0.0
episode_q_value = 0.0
episode_update_step = 0
episode_action_step = 0
time_episode_start = time.time()
game.begin_episode(steps_left)
while not game.episode_terminate:
# 1. We need to choose a new action based on the current game status
if game.state_enabled and game.replay_memory.sample_enabled:
do_exploration = (npy_rng.rand() < eps_curr)
eps_curr = max(eps_curr - eps_decay, eps_min)
if do_exploration:
action = npy_rng.randint(action_num)
else:
# TODO Here we can in fact play multiple gaming instances simultaneously and make actions for each
# We can simply stack the current_state() of gaming instances and give prediction for all of them
# We need to wait after calling calc_score(.), which makes the program slow
# TODO Profiling the speed of this part!
current_state = game.current_state()
state = nd.array(
current_state.reshape((1, ) + current_state.shape),
ctx=q_ctx) / float(255.0)
qval_npy = qnet.forward(is_train=False,
data=state)[0].asnumpy()
action = numpy.argmax(qval_npy)
episode_q_value += qval_npy[0, action]
episode_action_step += 1
else:
action = npy_rng.randint(action_num)
# 2. Play the game for a single mega-step (Inside the game, the action may be repeated for several times)
game.play(action)
total_steps += 1
# 3. Update our Q network if we can start sampling from the replay memory
# Also, we update every `update_interval`
if total_steps % update_interval == 0 and game.replay_memory.sample_enabled:
# 3.1 Draw sample from the replay_memory
training_steps += 1
episode_update_step += 1
states, actions, rewards, next_states, terminate_flags \
= game.replay_memory.sample(batch_size=minibatch_size)
states = nd.array(states, ctx=q_ctx) / float(255.0)
next_states = nd.array(next_states,
ctx=q_ctx) / float(255.0)
actions = nd.array(actions, ctx=q_ctx)
rewards = nd.array(rewards, ctx=q_ctx)
terminate_flags = nd.array(terminate_flags, ctx=q_ctx)
# 3.2 Use the target network to compute the scores and
# get the corresponding target rewards
if not args.double_q:
target_qval = target_qnet.forward(is_train=False,
data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(target_qval))\
* (1.0 - terminate_flags) * discount
else:
target_qval = target_qnet.forward(is_train=False,
data=next_states)[0]
qval = qnet.forward(is_train=False,
data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(qval))\
* (1.0 - terminate_flags) * discount
outputs = qnet.forward(is_train=True,
data=states,
dqn_action=actions,
dqn_reward=target_rewards)
qnet.backward()
if args.kv_type != None:
if use_easgd:
if total_steps % kvstore_update_period == 0:
for ind, k in enumerate(qnet.params.keys()):
kv.pull(ind,
central_weight[k],
priority=-ind)
qnet.params[k][:] -= easgd_alpha * \
(qnet.params[k] - central_weight[k])
kv.push(ind, qnet.params[k], priority=-ind)
qnet.update(updater=local_updater)
else:
update_on_kvstore(kv, qnet.params,
qnet.params_grad)
else:
qnet.update(updater=updater)
# 3.3 Calculate Loss
diff = nd.abs(
nd.choose_element_0index(outputs[0], actions) -
target_rewards)
quadratic_part = nd.clip(diff, -1, 1)
loss = 0.5 * nd.sum(nd.square(quadratic_part)).asnumpy()[0] +\
nd.sum(diff - quadratic_part).asnumpy()[0]
episode_loss += loss
# 3.3 Update the target network every freeze_interval
# (We can do annealing instead of hard copy)
if training_steps % freeze_interval == 0:
qnet.copy_params_to(target_qnet)
steps_left -= game.episode_step
time_episode_end = time.time()
# Update the statistics
epoch_reward += game.episode_reward
if args.kv_type != None:
info_str = "Node[%d]: " % kv.rank
else:
info_str = ""
info_str += "Epoch:%d, Episode:%d, Steps Left:%d/%d, Reward:%f, fps:%f, Exploration:%f" \
% (epoch, episode, steps_left, steps_per_epoch, game.episode_reward,
game.episode_step / (time_episode_end - time_episode_start), eps_curr)
if episode_update_step > 0:
info_str += ", Avg Loss:%f/%d" % (
episode_loss / episode_update_step, episode_update_step)
if episode_action_step > 0:
info_str += ", Avg Q Value:%f/%d" % (
episode_q_value / episode_action_step, episode_action_step)
logging.info(info_str)
end = time.time()
fps = steps_per_epoch / (end - start)
qnet.save_params(dir_path=args.dir_path, epoch=epoch)
if args.kv_type is not None:
logging.info(
"Node[%d]: Epoch:%d, FPS:%f, Avg Reward: %f/%d" %
(kv.rank, epoch, fps, epoch_reward / float(episode), episode))
else:
logging.info("Epoch:%d, FPS:%f, Avg Reward: %f/%d" %
(epoch, fps, epoch_reward / float(episode), episode))
def main():
parser = argparse.ArgumentParser(
description='Script to test the trained network on a game.')
parser.add_argument('-r',
'--rom',
required=False,
type=str,
default=os.path.join('arena', 'games', 'roms',
'breakout.bin'),
help='Path of the ROM File.')
parser.add_argument('-v',
'--visualization',
required=False,
type=int,
default=0,
help='Visualize the runs.')
parser.add_argument('--lr',
required=False,
type=float,
default=0.01,
help='Learning rate of the AdaGrad optimizer')
parser.add_argument('--eps',
required=False,
type=float,
default=0.01,
help='Eps of the AdaGrad optimizer')
parser.add_argument('--rms-decay',
required=False,
type=float,
default=0.95,
help='Decay rate of the RMSProp')
parser.add_argument('--clip-gradient',
required=False,
type=float,
default=None,
help='Clip threshold of the AdaGrad optimizer')
parser.add_argument('--double-q',
required=False,
type=bool,
default=False,
help='Use Double DQN')
parser.add_argument('--wd',
required=False,
type=float,
default=0.0,
help='Weight of the L2 Regularizer')
parser.add_argument(
'-c',
'--ctx',
required=False,
type=str,
default=None,
help='Running Context. E.g `-c gpu` or `-c gpu1` or `-c cpu`')
parser.add_argument('-d',
'--dir-path',
required=False,
type=str,
default='',
help='Saving directory of model files.')
parser.add_argument(
'--start-eps',
required=False,
type=float,
default=1.0,
help='Eps of the epsilon-greedy policy at the beginning')
parser.add_argument('--replay-start-size',
required=False,
type=int,
default=50000,
help='The step that the training starts')
parser.add_argument(
'--kvstore-update-period',
required=False,
type=int,
default=16,
help='The period that the worker updates the parameters from the sever'
)
parser.add_argument(
'--kv-type',
required=False,
type=str,
default=None,
help=
'type of kvstore, default will not use kvstore, could also be dist_async'
)
parser.add_argument('--optimizer',
required=False,
type=str,
default="adagrad",
help='type of optimizer')
parser.add_argument('--nactor',
required=False,
type=int,
default=16,
help='number of actor')
parser.add_argument('--exploration-period',
required=False,
type=int,
default=4000000,
help='length of annealing of epsilon greedy policy')
parser.add_argument('--replay-memory-size',
required=False,
type=int,
default=100,
help='size of replay memory')
parser.add_argument('--single-batch-size',
required=False,
type=int,
default=5,
help='batch size for every actor')
parser.add_argument('--symbol',
required=False,
type=str,
default="nature",
help='type of network, nature or nips')
parser.add_argument('--sample-policy',
required=False,
type=str,
default="recent",
help='minibatch sampling policy, recent or random')
parser.add_argument('--epoch-num',
required=False,
type=int,
default=50,
help='number of epochs')
parser.add_argument('--param-update-period',
required=False,
type=int,
default=5,
help='Parameter update period')
parser.add_argument('--resize-mode',
required=False,
type=str,
default="scale",
help='Resize mode, scale or crop')
parser.add_argument('--eps-update-period',
required=False,
type=int,
default=8000,
help='eps greedy policy update period')
parser.add_argument('--server-optimizer',
required=False,
type=str,
default="easgd",
help='type of server optimizer')
parser.add_argument('--nworker',
required=False,
type=int,
default=1,
help='number of kv worker')
parser.add_argument('--easgd-alpha',
required=False,
type=float,
default=0.01,
help='easgd alpha')
args, unknown = parser.parse_known_args()
logging.info(str(args))
if args.dir_path == '':
rom_name = os.path.splitext(os.path.basename(args.rom))[0]
time_str = time.strftime("%m%d_%H%M_%S", time.localtime())
args.dir_path = ('dqn-%s-%d_' % (rom_name,int(args.lr*10**5)))+time_str \
+ "_" + os.environ.get('DMLC_TASK_ID')
logging.info("saving to dir: " + args.dir_path)
if args.ctx == None:
args.ctx = os.environ.get('CTX')
logging.info("Context: %s" % args.ctx)
ctx = re.findall('([a-z]+)(\d*)', args.ctx)
ctx = [(device, int(num)) if len(num) > 0 else (device, 0)
for device, num in ctx]
# Async verision
nactor = args.nactor
param_update_period = args.param_update_period
replay_start_size = args.replay_start_size
max_start_nullops = 30
replay_memory_size = args.replay_memory_size
history_length = 4
rows = 84
cols = 84
q_ctx = mx.Context(*ctx[0])
games = []
for g in range(nactor):
games.append(
AtariGame(rom_path=args.rom,
resize_mode=args.resize_mode,
replay_start_size=replay_start_size,
resized_rows=rows,
resized_cols=cols,
max_null_op=max_start_nullops,
replay_memory_size=replay_memory_size,
display_screen=args.visualization,
history_length=history_length))
##RUN NATURE
freeze_interval = 40000 / nactor
freeze_interval /= param_update_period
epoch_num = args.epoch_num
steps_per_epoch = 4000000 / nactor
discount = 0.99
save_screens = False
eps_start = numpy.ones((3, )) * args.start_eps
eps_min = numpy.array([0.1, 0.01, 0.5])
eps_decay = (eps_start - eps_min) / (args.exploration_period / nactor)
eps_curr = eps_start
eps_id = numpy.zeros((nactor, ))
eps_update_period = args.eps_update_period
eps_update_count = numpy.zeros((nactor, ))
single_batch_size = args.single_batch_size
minibatch_size = nactor * single_batch_size
action_num = len(games[0].action_set)
data_shapes = {
'data': (minibatch_size, history_length) + (rows, cols),
'dqn_action': (minibatch_size, ),
'dqn_reward': (minibatch_size, )
}
if args.symbol == "nature":
dqn_sym = dqn_sym_nature(action_num)
elif args.symbol == "nips":
dqn_sym = dqn_sym_nips(action_num)
else:
raise NotImplementedError
qnet = Base(data_shapes=data_shapes,
sym=dqn_sym,
name='QNet',
initializer=DQNInitializer(factor_type="in"),
ctx=q_ctx)
target_qnet = qnet.copy(name="TargetQNet", ctx=q_ctx)
if args.optimizer == "adagrad":
optimizer = mx.optimizer.create(name=args.optimizer,
learning_rate=args.lr,
eps=args.eps,
clip_gradient=args.clip_gradient,
rescale_grad=1.0,
wd=args.wd)
elif args.optimizer == "rmsprop" or args.optimizer == "rmspropnoncentered":
optimizer = mx.optimizer.create(name=args.optimizer,
learning_rate=args.lr,
eps=args.eps,
clip_gradient=args.clip_gradient,
gamma1=args.rms_decay,
gamma2=0,
rescale_grad=1.0,
wd=args.wd)
lr_decay = (args.lr - 0) / (steps_per_epoch * epoch_num /
param_update_period)
# Create kvstore
use_easgd = False
if args.kv_type != None:
kvType = args.kv_type
kv = kvstore.create(kvType)
#Initialize kvstore
for idx, v in enumerate(qnet.params.values()):
kv.init(idx, v)
if args.server_optimizer == "easgd":
use_easgd = True
easgd_beta = 0.9
easgd_alpha = args.easgd_alpha
server_optimizer = mx.optimizer.create(name="ServerEasgd",
learning_rate=easgd_alpha)
easgd_eta = 0.00025
central_weight = OrderedDict([(n, v.copyto(q_ctx))
for n, v in qnet.params.items()])
kv.set_optimizer(server_optimizer)
updater = mx.optimizer.get_updater(optimizer)
else:
kv.set_optimizer(optimizer)
kvstore_update_period = args.kvstore_update_period
npy_rng = numpy.random.RandomState(123456 + kv.rank)
else:
updater = mx.optimizer.get_updater(optimizer)
qnet.print_stat()
target_qnet.print_stat()
states_buffer_for_act = numpy.zeros(
(nactor, history_length) + (rows, cols), dtype='uint8')
states_buffer_for_train = numpy.zeros(
(minibatch_size, history_length + 1) + (rows, cols), dtype='uint8')
next_states_buffer_for_train = numpy.zeros(
(minibatch_size, history_length) + (rows, cols), dtype='uint8')
actions_buffer_for_train = numpy.zeros((minibatch_size, ), dtype='uint8')
rewards_buffer_for_train = numpy.zeros((minibatch_size, ), dtype='float32')
terminate_flags_buffer_for_train = numpy.zeros((minibatch_size, ),
dtype='bool')
# Begin Playing Game
training_steps = 0
total_steps = 0
ave_fps = 0
ave_loss = 0
time_for_info = time.time()
parallel_executor = concurrent.futures.ThreadPoolExecutor(nactor)
for epoch in xrange(epoch_num):
# Run Epoch
steps_left = steps_per_epoch
episode = 0
epoch_reward = 0
start = time.time()
#
for g, game in enumerate(games):
game.start()
game.begin_episode()
eps_rand = npy_rng.rand()
if eps_rand < 0.4:
eps_id[g] = 0
elif eps_rand < 0.7:
eps_id[g] = 1
else:
eps_id[g] = 2
episode_stats = [EpisodeStat() for i in range(len(games))]
while steps_left > 0:
for g, game in enumerate(games):
if game.episode_terminate:
episode += 1
epoch_reward += game.episode_reward
if args.kv_type != None:
info_str = "Node[%d]: " % kv.rank
else:
info_str = ""
info_str += "Epoch:%d, Episode:%d, Steps Left:%d/%d, Reward:%f, fps:%f, Exploration:%f" \
% (epoch, episode, steps_left, steps_per_epoch, game.episode_reward,
ave_fps, (eps_curr[eps_id[g]]))
info_str += ", Avg Loss:%f" % ave_loss
if episode_stats[g].episode_action_step > 0:
info_str += ", Avg Q Value:%f/%d" % (
episode_stats[g].episode_q_value /
episode_stats[g].episode_action_step,
episode_stats[g].episode_action_step)
if g == 0: logging.info(info_str)
if eps_update_count[g] * eps_update_period < total_steps:
eps_rand = npy_rng.rand()
if eps_rand < 0.4:
eps_id[g] = 0
elif eps_rand < 0.7:
eps_id[g] = 1
else:
eps_id[g] = 2
eps_update_count[g] += 1
game.begin_episode(steps_left)
episode_stats[g] = EpisodeStat()
if total_steps > history_length:
for g, game in enumerate(games):
current_state = game.current_state()
states_buffer_for_act[g] = current_state
states = nd.array(states_buffer_for_act, ctx=q_ctx) / float(255.0)
qval_npy = qnet.forward(batch_size=nactor,
data=states)[0].asnumpy()
actions_that_max_q = numpy.argmax(qval_npy, axis=1)
actions = [0] * nactor
for g, game in enumerate(games):
# 1. We need to choose a new action based on the current game status
if games[g].state_enabled and games[
g].replay_memory.sample_enabled:
do_exploration = (npy_rng.rand() < eps_curr[eps_id[g]])
if do_exploration:
action = npy_rng.randint(action_num)
else:
# TODO Here we can in fact play multiple gaming instances simultaneously and make actions for each
# We can simply stack the current_state() of gaming instances and give prediction for all of them
# We need to wait after calling calc_score(.), which makes the program slow
# TODO Profiling the speed of this part!
action = actions_that_max_q[g]
episode_stats[g].episode_q_value += qval_npy[g, action]
episode_stats[g].episode_action_step += 1
else:
action = npy_rng.randint(action_num)
actions[g] = action
# t0=time.time()
for ret in parallel_executor.map(play_game, zip(games, actions)):
pass
# t1=time.time()
# logging.info("play time: %f" % (t1-t0))
eps_curr = numpy.maximum(eps_curr - eps_decay, eps_min)
total_steps += 1
steps_left -= 1
if total_steps % 100 == 0:
this_time = time.time()
ave_fps = (100 / (this_time - time_for_info))
time_for_info = this_time
# 3. Update our Q network if we can start sampling from the replay memory
# Also, we update every `update_interval`
if total_steps > minibatch_size and \
total_steps % (param_update_period) == 0 and \
games[-1].replay_memory.sample_enabled:
if use_easgd and training_steps % kvstore_update_period == 0:
for paramIndex in range(len(qnet.params)):
k = qnet.params.keys()[paramIndex]
kv.pull(paramIndex,
central_weight[k],
priority=-paramIndex)
qnet.params[k][:] -= easgd_alpha * (qnet.params[k] -
central_weight[k])
kv.push(paramIndex,
qnet.params[k],
priority=-paramIndex)
# 3.1 Draw sample from the replay_memory
for g, game in enumerate(games):
episode_stats[g].episode_update_step += 1
nsample = single_batch_size
i0 = (g * nsample)
i1 = (g + 1) * nsample
if args.sample_policy == "recent":
action, reward, terminate_flag=game.replay_memory.sample_last(batch_size=nsample,\
states=states_buffer_for_train,offset=i0)
elif args.sample_policy == "random":
action, reward, terminate_flag=game.replay_memory.sample_inplace(batch_size=nsample,\
states=states_buffer_for_train,offset=i0)
actions_buffer_for_train[i0:i1] = action
rewards_buffer_for_train[i0:i1] = reward
terminate_flags_buffer_for_train[i0:i1] = terminate_flag
states = nd.array(states_buffer_for_train[:, :-1],
ctx=q_ctx) / float(255.0)
next_states = nd.array(states_buffer_for_train[:, 1:],
ctx=q_ctx) / float(255.0)
actions = nd.array(actions_buffer_for_train, ctx=q_ctx)
rewards = nd.array(rewards_buffer_for_train, ctx=q_ctx)
terminate_flags = nd.array(terminate_flags_buffer_for_train,
ctx=q_ctx)
# 3.2 Use the target network to compute the scores and
# get the corresponding target rewards
if not args.double_q:
target_qval = target_qnet.forward(
batch_size=minibatch_size, data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(target_qval))\
* (1.0 - terminate_flags) * discount
else:
target_qval = target_qnet.forward(
batch_size=minibatch_size, data=next_states)[0]
qval = qnet.forward(batch_size=minibatch_size,
data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(qval))\
* (1.0 - terminate_flags) * discount
outputs = qnet.forward(batch_size=minibatch_size,
is_train=True,
data=states,
dqn_action=actions,
dqn_reward=target_rewards)
qnet.backward(batch_size=minibatch_size)
if args.kv_type is None or use_easgd:
qnet.update(updater=updater)
else:
update_on_kvstore(kv, qnet.params, qnet.params_grad)
# 3.3 Calculate Loss
diff = nd.abs(
nd.choose_element_0index(outputs[0], actions) -
target_rewards)
quadratic_part = nd.clip(diff, -1, 1)
loss = (0.5 * nd.sum(nd.square(quadratic_part)) +
nd.sum(diff - quadratic_part)).asscalar()
if ave_loss == 0:
ave_loss = loss
else:
ave_loss = 0.95 * ave_loss + 0.05 * loss
# 3.3 Update the target network every freeze_interval
# (We can do annealing instead of hard copy)
if training_steps % freeze_interval == 0:
qnet.copy_params_to(target_qnet)
if args.optimizer == "rmsprop" or args.optimizer == "rmspropnoncentered":
optimizer.lr -= lr_decay
if save_screens and training_steps % (
60 * 60 * 2 / param_update_period) == 0:
logging.info("saving screenshots")
for g in range(nactor):
screen = states_buffer_for_train[(
g * single_batch_size), -2, :, :].reshape(
states_buffer_for_train.shape[2:])
cv2.imwrite("screen_" + str(g) + ".png", screen)
training_steps += 1
end = time.time()
fps = steps_per_epoch / (end - start)
qnet.save_params(dir_path=args.dir_path, epoch=epoch)
if args.kv_type != None:
logging.info(
"Node[%d]: Epoch:%d, FPS:%f, Avg Reward: %f/%d" %
(kv.rank, epoch, fps, epoch_reward / float(episode), episode))
else:
logging.info("Epoch:%d, FPS:%f, Avg Reward: %f/%d" %
(epoch, fps, epoch_reward / float(episode), episode))
def main():
parser = argparse.ArgumentParser(
description='Script to test the trained network on a game.')
parser.add_argument('-r',
'--rom',
required=False,
type=str,
default=os.path.join('roms', 'breakout.bin'),
help='Path of the ROM File.')
parser.add_argument('-v',
'--visualization',
required=False,
type=int,
default=0,
help='Visualize the runs.')
parser.add_argument('--lr',
required=False,
type=float,
default=0.01,
help='Learning rate of the AdaGrad optimizer')
parser.add_argument('--eps',
required=False,
type=float,
default=0.01,
help='Eps of the AdaGrad optimizer')
parser.add_argument('--clip-gradient',
required=False,
type=float,
default=None,
help='Clip threshold of the AdaGrad optimizer')
parser.add_argument('--double-q',
required=False,
type=bool,
default=False,
help='Use Double DQN')
parser.add_argument('--wd',
required=False,
type=float,
default=0.0,
help='Weight of the L2 Regularizer')
parser.add_argument(
'-c',
'--ctx',
required=False,
type=str,
default='gpu',
help='Running Context. E.g `-c gpu` or `-c gpu1` or `-c cpu`')
parser.add_argument('-d',
'--dir-path',
required=False,
type=str,
default='',
help='Saving directory of model files.')
parser.add_argument(
'--start-eps',
required=False,
type=float,
default=1.0,
help='Eps of the epsilon-greedy policy at the beginning')
parser.add_argument('--replay-start-size',
required=False,
type=int,
default=50000,
help='The step that the training starts')
parser.add_argument(
'--kvstore-update-period',
required=False,
type=int,
default=1,
help='The period that the worker updates the parameters from the sever'
)
parser.add_argument(
'--kv-type',
required=False,
type=str,
default=None,
help=
'type of kvstore, default will not use kvstore, could also be dist_async'
)
parser.add_argument('--optimizer',
required=False,
type=str,
default="adagrad",
help='type of optimizer')
args = parser.parse_args()
## custom
args.start_eps = 0.2
args.replay_start_size = 1000
if args.dir_path == '':
rom_name = os.path.splitext(os.path.basename(args.rom))[0]
args.dir_path = 'dqn-%s-lr%g' % (rom_name, args.lr)
replay_start_size = args.replay_start_size
max_start_nullops = 30
replay_memory_size = 1000000
history_length = 4
rows = 84
cols = 84
ctx = parse_ctx(args.ctx)
q_ctx = mx.Context(*ctx[0])
game = AtariGame(rom_path=args.rom,
resize_mode='resize',
replay_start_size=replay_start_size,
resized_rows=rows,
resized_cols=cols,
max_null_op=max_start_nullops,
replay_memory_size=replay_memory_size,
display_screen=args.visualization,
history_length=history_length)
##RUN NATURE
freeze_interval = 10000
epoch_num = 30
steps_per_epoch = 100000
update_interval = 4
discount = 0.99
eps_start = args.start_eps
eps_min = 0.1
eps_decay = (eps_start - eps_min) / 1000000
eps_curr = eps_start
freeze_interval /= update_interval
minibatch_size = 32
action_num = len(game.action_set)
data_shapes = {
'data': (minibatch_size, history_length) + (rows, cols),
'dqn_action': (minibatch_size, ),
'dqn_reward': (minibatch_size, )
}
dqn_sym = dqn_sym_nature(action_num)
qnet = Base(data_shapes=data_shapes,
sym_gen=dqn_sym,
name='QNet',
initializer=DQNInitializer(factor_type="in"),
ctx=q_ctx)
target_qnet = qnet.copy(name="TargetQNet", ctx=q_ctx)
use_easgd = False
optimizer = mx.optimizer.create(name=args.optimizer,
learning_rate=args.lr,
eps=args.eps,
clip_gradient=args.clip_gradient,
rescale_grad=1.0,
wd=args.wd)
updater = mx.optimizer.get_updater(optimizer)
qnet.print_stat()
target_qnet.print_stat()
# Begin Playing Game
training_steps = 0
total_steps = 0
for epoch in range(epoch_num):
# Run Epoch
steps_left = steps_per_epoch
episode = 0
epoch_reward = 0
start = time.time()
game.start()
while steps_left > 0:
# Running New Episode
episode += 1
episode_loss = 0.0
episode_q_value = 0.0
episode_update_step = 0
episode_action_step = 0
time_episode_start = time.time()
game.begin_episode(steps_left)
while not game.episode_terminate:
# 1. We need to choose a new action based on the current game status
if game.state_enabled and game.replay_memory.sample_enabled:
do_exploration = (npy_rng.rand() < eps_curr)
eps_curr = max(eps_curr - eps_decay, eps_min)
if do_exploration:
action = npy_rng.randint(action_num)
else:
current_state = game.current_state()
state = nd.array(
current_state.reshape((1, ) + current_state.shape),
ctx=q_ctx) / float(255.0)
qval_npy = qnet.forward(is_train=False,
data=state)[0].asnumpy()
action = numpy.argmax(qval_npy)
episode_q_value += qval_npy[0, action]
episode_action_step += 1
else:
action = npy_rng.randint(action_num)
# 2. Play the game for a single mega-step (Inside the game, the action may be repeated for several times)
game.play(action)
total_steps += 1
# 3. Update our Q network if we can start sampling from the replay memory
# Also, we update every `update_interval`
if total_steps % update_interval == 0 and game.replay_memory.sample_enabled:
# 3.1 Draw sample from the replay_memory
training_steps += 1
episode_update_step += 1
states, actions, rewards, next_states, terminate_flags \
= game.replay_memory.sample(batch_size=minibatch_size)
states = nd.array(states, ctx=q_ctx) / float(255.0)
next_states = nd.array(next_states,
ctx=q_ctx) / float(255.0)
actions = nd.array(actions, ctx=q_ctx)
rewards = nd.array(rewards, ctx=q_ctx)
terminate_flags = nd.array(terminate_flags, ctx=q_ctx)
# 3.2 Use the target network to compute the scores and
# get the corresponding target rewards
if not args.double_q:
target_qval = target_qnet.forward(is_train=False,
data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(target_qval)) \
* (1.0 - terminate_flags) * discount
else:
target_qval = target_qnet.forward(is_train=False,
data=next_states)[0]
qval = qnet.forward(is_train=False,
data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(qval)) \
* (1.0 - terminate_flags) * discount
outputs = qnet.forward(is_train=True,
data=states,
dqn_action=actions,
dqn_reward=target_rewards)
qnet.backward()
qnet.update(updater=updater)
# 3.3 Calculate Loss
diff = nd.abs(
nd.choose_element_0index(outputs[0], actions) -
target_rewards)
quadratic_part = nd.clip(diff, -1, 1)
loss = 0.5 * nd.sum(nd.square(quadratic_part)).asnumpy()[0] + \
nd.sum(diff - quadratic_part).asnumpy()[0]
episode_loss += loss
# 3.3 Update the target network every freeze_interval
if training_steps % freeze_interval == 0:
qnet.copy_params_to(target_qnet)
steps_left -= game.episode_step
time_episode_end = time.time()
# Update the statistics
epoch_reward += game.episode_reward
info_str = "Epoch:%d, Episode:%d, Steps Left:%d/%d, Reward:%f, fps:%f, Exploration:%f" \
% (epoch, episode, steps_left, steps_per_epoch, game.episode_reward,
game.episode_step / (time_episode_end - time_episode_start), eps_curr)
if episode_update_step > 0:
info_str += ", Avg Loss:%f/%d" % (
episode_loss / episode_update_step, episode_update_step)
if episode_action_step > 0:
info_str += ", Avg Q Value:%f/%d" % (
episode_q_value / episode_action_step, episode_action_step)
if episode % 100 == 0:
logging.info(info_str)
end = time.time()
fps = steps_per_epoch / (end - start)
# qnet.save_params(dir_path=args.dir_path, epoch=epoch)
logging.info("Epoch:%d, FPS:%f, Avg Reward: %f/%d" %
(epoch, fps, epoch_reward / float(episode), episode))
def main():
parser = argparse.ArgumentParser(description='Script to test the trained network on a game.')
parser.add_argument('-r', '--rom', required=False, type=str,
default=os.path.join('arena', 'games', 'roms', 'breakout.bin'),
help='Path of the ROM File.')
parser.add_argument('-v', '--visualization', required=False, type=int, default=0,
help='Visualize the runs.')
parser.add_argument('--lr', required=False, type=float, default=0.01,
help='Learning rate of the AdaGrad optimizer')
parser.add_argument('--eps', required=False, type=float, default=0.01,
help='Eps of the AdaGrad optimizer')
parser.add_argument('--clip-gradient', required=False, type=float, default=None,
help='Clip threshold of the AdaGrad optimizer')
parser.add_argument('--double-q', required=False, type=bool, default=False,
help='Use Double DQN')
parser.add_argument('--wd', required=False, type=float, default=0.0,
help='Weight of the L2 Regularizer')
parser.add_argument('-c', '--ctx', required=False, type=str, default='gpu',
help='Running Context. E.g `-c gpu` or `-c gpu1` or `-c cpu`')
parser.add_argument('-d', '--dir-path', required=False, type=str, default='',
help='Saving directory of model files.')
parser.add_argument('--start-eps', required=False, type=float, default=1.0,
help='Eps of the epsilon-greedy policy at the beginning')
parser.add_argument('--replay-start-size', required=False, type=int, default=50000,
help='The step that the training starts')
parser.add_argument('--kvstore-update-period', required=False, type=int, default=1,
help='The period that the worker updates the parameters from the sever')
parser.add_argument('--kv-type', required=False, type=str, default=None,
help='type of kvstore, default will not use kvstore, could also be dist_async')
args, unknown = parser.parse_known_args()
if args.dir_path == '':
rom_name = os.path.splitext(os.path.basename(args.rom))[0]
args.dir_path = 'dqn-%s' % rom_name
ctx = re.findall('([a-z]+)(\d*)', args.ctx)
ctx = [(device, int(num)) if len(num) >0 else (device, 0) for device, num in ctx]
replay_start_size = args.replay_start_size
max_start_nullops = 30
replay_memory_size = 1000000
history_length = 4
rows = 84
cols = 84
q_ctx = mx.Context(*ctx[0])
game = AtariGame(rom_path=args.rom, resize_mode='scale', replay_start_size=replay_start_size,
resized_rows=rows, resized_cols=cols, max_null_op=max_start_nullops,
replay_memory_size=replay_memory_size, display_screen=args.visualization,
history_length=history_length)
##RUN NATURE
freeze_interval = 10000
epoch_num = 200
steps_per_epoch = 250000
update_interval = 4
discount = 0.99
eps_start = args.start_eps
eps_min = 0.1
eps_decay = (eps_start - 0.1) / 1000000
eps_curr = eps_start
freeze_interval /= update_interval
minibatch_size = 32
action_num = len(game.action_set)
data_shapes = {'data': (minibatch_size, history_length) + (rows, cols),
'dqn_action': (minibatch_size,), 'dqn_reward': (minibatch_size,)}
#optimizer = mx.optimizer.create(name='sgd', learning_rate=args.lr,wd=args.wd)
optimizer = mx.optimizer.Nop()
dqn_output_op = DQNOutputNpyOp()
dqn_sym = dqn_sym_nature(action_num, dqn_output_op)
qnet = Base(data_shapes=data_shapes, sym=dqn_sym, name='QNet',
initializer=DQNInitializer(factor_type="in"),
ctx=q_ctx)
target_qnet = qnet.copy(name="TargetQNet", ctx=q_ctx)
# Create kvstore
testShape = (1,1686180*100)
testParam = nd.ones(testShape,ctx=q_ctx)
testGrad = nd.zeros(testShape,ctx=q_ctx)
# Create kvstore
if args.kv_type != None:
kvType = args.kv_type
kvStore = kvstore.create(kvType)
#Initialize kvstore
for idx,v in enumerate(qnet.params.values()):
kvStore.init(idx,v);
# Set optimizer on kvstore
kvStore.set_optimizer(optimizer)
kvstore_update_period = args.kvstore_update_period
else:
updater = mx.optimizer.get_updater(optimizer)
# if args.kv_type != None:
# kvType = args.kv_type
# kvStore = kvstore.create(kvType)
# kvStore.init(0,testParam)
# testOptimizer = mx.optimizer.Nop()
# kvStore.set_optimizer(testOptimizer)
# kvstore_update_period = args.kvstore_update_period
qnet.print_stat()
target_qnet.print_stat()
# Begin Playing Game
training_steps = 0
total_steps = 0
while(1):
time_before_wait = time.time()
# kvStore.push(0,testGrad,priority=0)
# kvStore.pull(0,testParam,priority=0)
# testParam.wait_to_read()
for paramIndex in range(len(qnet.params)):#range(6):#
k=qnet.params.keys()[paramIndex]
kvStore.push(paramIndex,qnet.params_grad[k],priority=-paramIndex)
kvStore.pull(paramIndex,qnet.params[k],priority=-paramIndex)
for v in qnet.params.values():
v.wait_to_read()
logging.info("wait time %f" %(time.time()-time_before_wait))
for epoch in xrange(epoch_num):
# Run Epoch
steps_left = steps_per_epoch
episode = 0
epoch_reward = 0
start = time.time()
game.start()
while steps_left > 0:
# Running New Episode
episode += 1
episode_loss = 0.0
episode_q_value = 0.0
episode_update_step = 0
episode_action_step = 0
time_episode_start = time.time()
game.begin_episode(steps_left)
while not game.episode_terminate:
# 1. We need to choose a new action based on the current game status
if game.state_enabled and game.replay_memory.sample_enabled:
do_exploration = (npy_rng.rand() < eps_curr)
eps_curr = max(eps_curr - eps_decay, eps_min)
if do_exploration:
action = npy_rng.randint(action_num)
else:
# TODO Here we can in fact play multiple gaming instances simultaneously and make actions for each
# We can simply stack the current_state() of gaming instances and give prediction for all of them
# We need to wait after calling calc_score(.), which makes the program slow
# TODO Profiling the speed of this part!
current_state = game.current_state()
state = nd.array(current_state.reshape((1,) + current_state.shape),
ctx=q_ctx) / float(255.0)
qval_npy = qnet.forward(batch_size=1, data=state)[0].asnumpy()
action = numpy.argmax(qval_npy)
episode_q_value += qval_npy[0, action]
episode_action_step += 1
else:
action = npy_rng.randint(action_num)
# 2. Play the game for a single mega-step (Inside the game, the action may be repeated for several times)
game.play(action)
total_steps += 1
# 3. Update our Q network if we can start sampling from the replay memory
# Also, we update every `update_interval`
if total_steps % update_interval == 0 and game.replay_memory.sample_enabled:
# 3.1 Draw sample from the replay_memory
training_steps += 1
episode_update_step += 1
states, actions, rewards, next_states, terminate_flags \
= game.replay_memory.sample(batch_size=minibatch_size)
states = nd.array(states, ctx=q_ctx) / float(255.0)
next_states = nd.array(next_states, ctx=q_ctx) / float(255.0)
actions = nd.array(actions, ctx=q_ctx)
rewards = nd.array(rewards, ctx=q_ctx)
terminate_flags = nd.array(terminate_flags, ctx=q_ctx)
# 3.2 Use the target network to compute the scores and
# get the corresponding target rewards
if not args.double_q:
target_qval = target_qnet.forward(batch_size=minibatch_size,
data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(target_qval))\
* (1.0 - terminate_flags) * discount
else:
target_qval = target_qnet.forward(batch_size=minibatch_size,
data=next_states)[0]
qval = qnet.forward(batch_size=minibatch_size, data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(qval))\
* (1.0 - terminate_flags) * discount
outputs = qnet.forward(batch_size=minibatch_size,is_train=True, data=states,
dqn_action=actions,
dqn_reward=target_rewards)
qnet.backward(batch_size=minibatch_size)
nd.waitall()
time_before_update = time.time()
if args.kv_type != None:
if total_steps % kvstore_update_period == 0:
update_to_kvstore(kvStore,qnet.params,qnet.params_grad)
else:
qnet.update(updater=updater)
logging.info("update time %f" %(time.time()-time_before_update))
time_before_wait = time.time()
nd.waitall()
logging.info("wait time %f" %(time.time()-time_before_wait))
'''nd.waitall()
time_before_wait = time.time()
kvStore.push(0,testGrad,priority=0)
kvStore.pull(0,testParam,priority=0)
nd.waitall()
logging.info("wait time %f" %(time.time()-time_before_wait))'''
# 3.3 Calculate Loss
diff = nd.abs(nd.choose_element_0index(outputs[0], actions) - target_rewards)
quadratic_part = nd.clip(diff, -1, 1)
loss = (0.5 * nd.sum(nd.square(quadratic_part)) + nd.sum(diff - quadratic_part)).asscalar()
episode_loss += loss
# 3.3 Update the target network every freeze_interval
# (We can do annealing instead of hard copy)
if training_steps % freeze_interval == 0:
qnet.copy_params_to(target_qnet)
steps_left -= game.episode_step
time_episode_end = time.time()
# Update the statistics
epoch_reward += game.episode_reward
info_str = "Epoch:%d, Episode:%d, Steps Left:%d/%d, Reward:%f, fps:%f, Exploration:%f" \
% (epoch, episode, steps_left, steps_per_epoch, game.episode_reward,
game.episode_step / (time_episode_end - time_episode_start), eps_curr)
if episode_update_step > 0:
info_str += ", Avg Loss:%f/%d" % (episode_loss / episode_update_step,
episode_update_step)
if episode_action_step > 0:
info_str += ", Avg Q Value:%f/%d" % (episode_q_value / episode_action_step,
episode_action_step)
logging.info(info_str)
end = time.time()
fps = steps_per_epoch / (end - start)
qnet.save_params(dir_path=args.dir_path, epoch=epoch)
logging.info("Epoch:%d, FPS:%f, Avg Reward: %f/%d"
% (epoch, fps, epoch_reward / float(episode), episode))
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--num-envs', type=int, default=1)
parser.add_argument('--t-max', type=int, default=1)
parser.add_argument('--learning-rate', type=float, default=0.0002)
parser.add_argument('--seed', type=int, default=0)
parser.add_argument('--steps-per-epoch', type=int, default=100000)
parser.add_argument('--testing', type=int, default=0)
parser.add_argument('--continue-training', type=int, default=0)
parser.add_argument('--epoch-num', type=int, default=40)
parser.add_argument('--start-epoch', type=int, default=20)
parser.add_argument('--testing-epoch', type=int, default=3)
parser.add_argument('--save-log', type=str, default='basic/log')
parser.add_argument('--signal-num', type=int, default=4)
parser.add_argument('--toxin', type=int, default=0)
parser.add_argument('--a1-AC-folder', type=str, default='basic/a1_Qnet')
parser.add_argument('--eps-start', type=float, default=1.0)
parser.add_argument('--replay-start-size', type=int, default=50000)
parser.add_argument('--decay-rate', type=int, default=500000)
parser.add_argument('--replay-memory-size', type=int, default=1000000)
parser.add_argument('--eps-min', type=float, default=0.05)
rewards = {
"positive": 1.0,
"negative": -1.0,
"tick": -0.002,
"loss": -2.0,
"win": 2.0
}
args = parser.parse_args()
config = Config(args)
q_ctx = config.ctx
steps_per_epoch = args.steps_per_epoch
np.random.seed(args.seed)
start_epoch = args.start_epoch
testing_epoch = args.testing_epoch
save_log = args.save_log
epoch_num = args.epoch_num
epoch_range = range(epoch_num)
toxin = args.toxin
a1_Qnet_folder = args.a1_AC_folder
freeze_interval = 10000
update_interval = 5
replay_memory_size = args.replay_memory_size
discount = 0.99
replay_start_size = args.replay_start_size
history_length = 1
eps_start = args.eps_start
eps_min = args.eps_min
eps_decay = (eps_start - eps_min) / args.decay_rate
eps_curr = eps_start
freeze_interval /= update_interval
minibatch_size = 32
testing = args.testing
testing = True if testing == 1 else False
continue_training = args.continue_training
continue_training = True if continue_training == 1 else False
game = HunterWorld(width=256,
height=256,
num_preys=10,
draw=False,
num_hunters=2,
num_toxins=toxin)
env = PLE(game,
fps=30,
force_fps=True,
display_screen=False,
reward_values=rewards,
resized_rows=80,
resized_cols=80,
num_steps=2)
replay_memory = ReplayMemory(state_dim=(148, ),
history_length=history_length,
memory_size=replay_memory_size,
replay_start_size=replay_start_size,
state_dtype='float32')
action_set = env.get_action_set()
action_map = []
for action1 in action_set[0].values():
for action2 in action_set[1].values():
action_map.append([action1, action2])
action_map = np.array(action_map)
action_num = action_map.shape[0]
target1 = Qnetwork(actions_num=action_num,
q_ctx=q_ctx,
isTrain=False,
batch_size=1,
dir=dir,
folder=a1_Qnet_folder)
target32 = Qnetwork(actions_num=action_num,
q_ctx=q_ctx,
isTrain=False,
batch_size=32,
dir=dir,
folder=a1_Qnet_folder)
Qnet = Qnetwork(actions_num=action_num,
q_ctx=q_ctx,
isTrain=True,
batch_size=32,
dir=dir,
folder=a1_Qnet_folder)
if testing:
env.force_fps = False
env.game.draw = True
env.display_screen = True
Qnet.load_params(testing_epoch)
elif continue_training:
epoch_range = range(start_epoch, epoch_num + start_epoch)
Qnet.load_params(start_epoch - 1)
logging_config(logging, dir, save_log, file_name)
else:
logging_config(logging, dir, save_log, file_name)
copyTargetQNetwork(Qnet.model, target1.model)
copyTargetQNetwork(Qnet.model, target32.model)
logging.info('args=%s' % args)
logging.info('config=%s' % config.__dict__)
print_params(logging, Qnet.model)
training_steps = 0
total_steps = 0
for epoch in epoch_range:
steps_left = steps_per_epoch
episode = 0
epoch_reward = 0
start = time.time()
env.reset_game()
while steps_left > 0:
episode += 1
episode_loss = 0.0
episode_q_value = 0.0
episode_update_step = 0
episode_action_step = 0
episode_reward = 0
episode_step = 0
collisions = 0.0
time_episode_start = time.time()
env.reset_game()
while not env.game_over():
if replay_memory.size >= history_length and replay_memory.size > replay_start_size:
do_exploration = (np.random.rand() < eps_curr)
eps_curr = max(eps_curr - eps_decay, eps_min)
if do_exploration:
action = np.random.randint(action_num)
else:
current_state = replay_memory.latest_slice()
state = nd.array(
current_state.reshape((1, ) + current_state.shape),
ctx=q_ctx)
target1.model.forward(mx.io.DataBatch([state], []))
q_value = target1.model.get_outputs()[0].asnumpy()[0]
action = numpy.argmax(q_value)
episode_q_value += q_value[action]
episode_action_step += 1
else:
action = np.random.randint(action_num)
next_ob, reward, terminal_flag = env.act(action_map[action])
reward = np.sum(reward)
replay_memory.append(
np.array(next_ob).flatten(), action, reward, terminal_flag)
total_steps += 1
episode_reward += reward
if reward < 0:
collisions += 1
episode_step += 1
if total_steps % update_interval == 0 and replay_memory.size > replay_start_size:
training_steps += 1
state_batch, actions, rewards, nextstate_batch, terminate_flags = replay_memory.sample(
batch_size=minibatch_size)
state_batch = nd.array(state_batch, ctx=q_ctx)
actions_batch = nd.array(actions, ctx=q_ctx)
reward_batch = nd.array(rewards, ctx=q_ctx)
terminate_flags = nd.array(terminate_flags, ctx=q_ctx)
target32.model.forward(
mx.io.DataBatch([nd.array(nextstate_batch, ctx=q_ctx)],
[]))
Qvalue = target32.model.get_outputs()[0]
y_batch = reward_batch + nd.choose_element_0index(
Qvalue, nd.argmax_channel(Qvalue)) * (
1.0 - terminate_flags) * discount
Qnet.model.forward(mx.io.DataBatch(
[state_batch, actions_batch, y_batch], []),
is_train=True)
Qnet.model.backward()
Qnet.model.update()
if training_steps % 10 == 0:
loss1 = 0.5 * nd.square(
nd.choose_element_0index(
Qnet.model.get_outputs()[0], actions_batch) -
y_batch)
episode_loss += nd.sum(loss1).asnumpy()
episode_update_step += 1
if training_steps % freeze_interval == 0:
copyTargetQNetwork(Qnet.model, target1.model)
copyTargetQNetwork(Qnet.model, target32.model)
steps_left -= episode_step
time_episode_end = time.time()
epoch_reward += episode_reward
info_str = "Epoch:%d, Episode:%d, Steps Left:%d/%d/%d, Reward:%f, fps:%f, Exploration:%f" \
% (epoch, episode, steps_left, episode_step, steps_per_epoch, episode_reward,
episode_step / (time_episode_end - time_episode_start), eps_curr)
info_str += ", Collision:%f/%d " % (collisions / episode_step,
collisions)
if episode_update_step > 0:
info_str += ", Avg Loss:%f/%d" % (
episode_loss / episode_update_step, episode_update_step)
if episode_action_step > 0:
info_str += ", Avg Q Value:%f/%d " % (
episode_q_value / episode_action_step, episode_action_step)
if episode % 1 == 0:
logging.info(info_str)
print info_str
end = time.time()
fps = steps_per_epoch / (end - start)
Qnet.save_params(epoch)
print "Epoch:%d, FPS:%f, Avg Reward: %f/%d" % (
epoch, fps, epoch_reward / float(episode), episode)
def main():
parser = argparse.ArgumentParser(description='Script to test the trained network on a game.')
parser.add_argument('-r', '--rom', required=False, type=str,
default=os.path.join('roms', 'breakout.bin'),
help='Path of the ROM File.')
parser.add_argument('-v', '--visualization', action='store_true',
help='Visualize the runs.')
parser.add_argument('--lr', required=False, type=float, default=0.01,
help='Learning rate of the AdaGrad optimizer')
parser.add_argument('--eps', required=False, type=float, default=0.01,
help='Eps of the AdaGrad optimizer')
parser.add_argument('--clip-gradient', required=False, type=float, default=None,
help='Clip threshold of the AdaGrad optimizer')
parser.add_argument('--double-q', action='store_true',
help='Use Double DQN only if specified')
parser.add_argument('--wd', required=False, type=float, default=0.0,
help='Weight of the L2 Regularizer')
parser.add_argument('-c', '--ctx', required=False, type=str, default='gpu',
help='Running Context. E.g `-c gpu` or `-c gpu1` or `-c cpu`')
parser.add_argument('-d', '--dir-path', required=False, type=str, default='',
help='Saving directory of model files.')
parser.add_argument('--start-eps', required=False, type=float, default=1.0,
help='Eps of the epsilon-greedy policy at the beginning')
parser.add_argument('--replay-start-size', required=False, type=int, default=50000,
help='The step that the training starts')
parser.add_argument('--kvstore-update-period', required=False, type=int, default=1,
help='The period that the worker updates the parameters from the sever')
parser.add_argument('--kv-type', required=False, type=str, default=None,
help='type of kvstore, default will not use kvstore, could also be dist_async')
parser.add_argument('--optimizer', required=False, type=str, default="adagrad",
help='type of optimizer')
args = parser.parse_args()
if args.dir_path == '':
rom_name = os.path.splitext(os.path.basename(args.rom))[0]
args.dir_path = 'dqn-%s-lr%g' % (rom_name, args.lr)
replay_start_size = args.replay_start_size
max_start_nullops = 30
replay_memory_size = 1000000
history_length = 4
rows = 84
cols = 84
ctx = parse_ctx(args.ctx)
q_ctx = mx.Context(*ctx[0])
game = AtariGame(rom_path=args.rom, resize_mode='scale', replay_start_size=replay_start_size,
resized_rows=rows, resized_cols=cols, max_null_op=max_start_nullops,
replay_memory_size=replay_memory_size, display_screen=args.visualization,
history_length=history_length)
##RUN NATURE
freeze_interval = 10000
epoch_num = 200
steps_per_epoch = 250000
update_interval = 4
discount = 0.99
eps_start = args.start_eps
eps_min = 0.1
eps_decay = (eps_start - eps_min) / 1000000
eps_curr = eps_start
freeze_interval /= update_interval
minibatch_size = 32
action_num = len(game.action_set)
data_shapes = {'data': (minibatch_size, history_length) + (rows, cols),
'dqn_action': (minibatch_size,), 'dqn_reward': (minibatch_size,)}
dqn_sym = dqn_sym_nature(action_num)
qnet = Base(data_shapes=data_shapes, sym_gen=dqn_sym, name='QNet',
initializer=DQNInitializer(factor_type="in"),
ctx=q_ctx)
target_qnet = qnet.copy(name="TargetQNet", ctx=q_ctx)
use_easgd = False
optimizer = mx.optimizer.create(name=args.optimizer, learning_rate=args.lr, eps=args.eps,
clip_gradient=args.clip_gradient,
rescale_grad=1.0, wd=args.wd)
updater = mx.optimizer.get_updater(optimizer)
qnet.print_stat()
target_qnet.print_stat()
# Begin Playing Game
training_steps = 0
total_steps = 0
for epoch in range(epoch_num):
# Run Epoch
steps_left = steps_per_epoch
episode = 0
epoch_reward = 0
start = time.time()
game.start()
while steps_left > 0:
# Running New Episode
episode += 1
episode_loss = 0.0
episode_q_value = 0.0
episode_update_step = 0
episode_action_step = 0
time_episode_start = time.time()
game.begin_episode(steps_left)
while not game.episode_terminate:
# 1. We need to choose a new action based on the current game status
if game.state_enabled and game.replay_memory.sample_enabled:
do_exploration = (npy_rng.rand() < eps_curr)
eps_curr = max(eps_curr - eps_decay, eps_min)
if do_exploration:
action = npy_rng.randint(action_num)
else:
# TODO Here we can in fact play multiple gaming instances simultaneously and make actions for each
# We can simply stack the current_state() of gaming instances and give prediction for all of them
# We need to wait after calling calc_score(.), which makes the program slow
# TODO Profiling the speed of this part!
current_state = game.current_state()
state = nd.array(current_state.reshape((1,) + current_state.shape),
ctx=q_ctx) / float(255.0)
qval_npy = qnet.forward(is_train=False, data=state)[0].asnumpy()
action = numpy.argmax(qval_npy)
episode_q_value += qval_npy[0, action]
episode_action_step += 1
else:
action = npy_rng.randint(action_num)
# 2. Play the game for a single mega-step (Inside the game, the action may be repeated for several times)
game.play(action)
total_steps += 1
# 3. Update our Q network if we can start sampling from the replay memory
# Also, we update every `update_interval`
if total_steps % update_interval == 0 and game.replay_memory.sample_enabled:
# 3.1 Draw sample from the replay_memory
training_steps += 1
episode_update_step += 1
states, actions, rewards, next_states, terminate_flags \
= game.replay_memory.sample(batch_size=minibatch_size)
states = nd.array(states, ctx=q_ctx) / float(255.0)
next_states = nd.array(next_states, ctx=q_ctx) / float(255.0)
actions = nd.array(actions, ctx=q_ctx)
rewards = nd.array(rewards, ctx=q_ctx)
terminate_flags = nd.array(terminate_flags, ctx=q_ctx)
# 3.2 Use the target network to compute the scores and
# get the corresponding target rewards
if not args.double_q:
target_qval = target_qnet.forward(is_train=False, data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(target_qval))\
* (1.0 - terminate_flags) * discount
else:
target_qval = target_qnet.forward(is_train=False, data=next_states)[0]
qval = qnet.forward(is_train=False, data=next_states)[0]
target_rewards = rewards + nd.choose_element_0index(target_qval,
nd.argmax_channel(qval))\
* (1.0 - terminate_flags) * discount
outputs = qnet.forward(is_train=True,
data=states,
dqn_action=actions,
dqn_reward=target_rewards)
qnet.backward()
qnet.update(updater=updater)
# 3.3 Calculate Loss
diff = nd.abs(nd.choose_element_0index(outputs[0], actions) - target_rewards)
quadratic_part = nd.clip(diff, -1, 1)
loss = 0.5 * nd.sum(nd.square(quadratic_part)).asnumpy()[0] +\
nd.sum(diff - quadratic_part).asnumpy()[0]
episode_loss += loss
# 3.3 Update the target network every freeze_interval
if training_steps % freeze_interval == 0:
qnet.copy_params_to(target_qnet)
steps_left -= game.episode_step
time_episode_end = time.time()
# Update the statistics
epoch_reward += game.episode_reward
info_str = "Epoch:%d, Episode:%d, Steps Left:%d/%d, Reward:%f, fps:%f, Exploration:%f" \
% (epoch, episode, steps_left, steps_per_epoch, game.episode_reward,
game.episode_step / (time_episode_end - time_episode_start), eps_curr)
if episode_update_step > 0:
info_str += ", Avg Loss:%f/%d" % (episode_loss / episode_update_step,
episode_update_step)
if episode_action_step > 0:
info_str += ", Avg Q Value:%f/%d" % (episode_q_value / episode_action_step,
episode_action_step)
if episode % 100 == 0:
logging.info(info_str)
end = time.time()
fps = steps_per_epoch / (end - start)
qnet.save_params(dir_path=args.dir_path, epoch=epoch)
logging.info("Epoch:%d, FPS:%f, Avg Reward: %f/%d"
% (epoch, fps, epoch_reward / float(episode), episode))
print q_net.calc_score(
batch_size=minibatch_size,
data=numpy.random.normal(
0, 1, data_shapes['data']).astype('float32'))[0].asnumpy()
q_net.print_stat()
target_q_net.print_stat()
sample = numpy.random.normal(0, 1, data_shapes['data']).astype('float32')
start = time.time()
for i in xrange(100):
action = numpy.random.randint(low=0, high=4, size=(minibatch_size, ))
target_action = nd.array(action, ctx=mx.gpu())
target_scores = target_q_net.calc_score(batch_size=minibatch_size,
data=sample)[0]
target_reward = nd.choose_element_0index(target_scores, target_action)
q_net.fit_target(batch_size=minibatch_size,
data=sample,
dqn_action=target_action,
dqn_reward=target_reward)
predict_reward = nd.choose_element_0index(
q_net.calc_score(batch_size=minibatch_size, data=sample)[0],
target_action)
another_predict_reward = nd.choose_element_0index(
another_target_q_net.calc_score(batch_size=minibatch_size,
data=sample)[0], target_action)
print nd.sum(nd.square(target_reward - predict_reward)).asscalar(), nd.sum(
nd.square(another_predict_reward - predict_reward)).asscalar()
end = time.time()
print end - start