gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-17 19:20:19 +01:00
Code Issues Projects Releases Wiki Activity

Cleanup imports.

Browse Source 
Till now, most of the modules were importing all of the module objects
(variables, classes, functions, other imports) into module namespace,
which potentially could (and was) cause of unintentional use of class or
methods, which was indirect imported.

With this patch, all the star imports were substituted with top-level
module, which provides desired class or function.

Besides, all imports where sorted (where possible) in a way pep8[1]
suggests - first are imports from standard library, than goes third
party imports (like numpy, tensorflow etc) and finally coach modules.
All of those sections are separated by one empty line.

[1] https://www.python.org/dev/peps/pep-0008/#imports
This commit is contained in:
Roman Dobosz
2018-04-12 19:46:32 +02:00
parent cafa152382
commit 1b095aeeca
75 changed files with 1169 additions and 1139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
agents/ddpg_agent.py
View File

@@ -13,28 +13,34 @@
# See the License for the specific language governing permissions and
# limitations under the License.
#
import copy
from agents.actor_critic_agent import *
from configurations import *
import numpy as np
from agents import actor_critic_agent as aca
from agents import agent
from architectures import network_wrapper as nw
import configurations as conf
import utils
# Deep Deterministic Policy Gradients Network - https://arxiv.org/pdf/1509.02971.pdf
class DDPGAgent(ActorCriticAgent):
class DDPGAgent(aca.ActorCriticAgent):
def __init__(self, env, tuning_parameters, replicated_device=None, thread_id=0):
ActorCriticAgent.__init__(self, env, tuning_parameters, replicated_device, thread_id,
create_target_network=True)
aca.ActorCriticAgent.__init__(self, env, tuning_parameters, replicated_device, thread_id,
create_target_network=True)
# define critic network
self.critic_network = self.main_network
# self.networks.append(self.critic_network)
# define actor network
tuning_parameters.agent.input_types = {'observation': InputTypes.Observation}
tuning_parameters.agent.output_types = [OutputTypes.Pi]
self.actor_network = NetworkWrapper(tuning_parameters, True, self.has_global, 'actor',
self.replicated_device, self.worker_device)
tuning_parameters.agent.input_types = {'observation': conf.InputTypes.Observation}
tuning_parameters.agent.output_types = [conf.OutputTypes.Pi]
self.actor_network = nw.NetworkWrapper(tuning_parameters, True, self.has_global, 'actor',
self.replicated_device, self.worker_device)
self.networks.append(self.actor_network)
self.q_values = Signal("Q")
self.q_values = utils.Signal("Q")
self.signals.append(self.q_values)
self.reset_game(do_not_reset_env=True)
@@ -82,14 +88,14 @@ class DDPGAgent(ActorCriticAgent):
return total_loss
def train(self):
return Agent.train(self)
return agent.Agent.train(self)
def choose_action(self, curr_state, phase=RunPhase.TRAIN):
def choose_action(self, curr_state, phase=utils.RunPhase.TRAIN):
assert not self.env.discrete_controls, 'DDPG works only for continuous control problems'
result = self.actor_network.online_network.predict(self.tf_input_state(curr_state))
action_values = result[0].squeeze()
if phase == RunPhase.TRAIN:
if phase == utils.RunPhase.TRAIN:
action = self.exploration_policy.get_action(action_values)
else:
action = action_values