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

create method LevelManager.acting_agent()

Browse Source
This commit is contained in:
zach dwiel
2019-04-05 10:42:48 -04:00
committed by Zach Dwiel
parent 54fdfe2da8
commit b20e795ce0
1 changed files with 9 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
rl_coach/level_manager.py
View File

@@ -205,6 +205,13 @@ class LevelManager(EnvironmentInterface):
for agent in self.agents.values(): for agent in self.agents.values():
agent.phase = val agent.phase = val
def acting_agent(self) -> AgentInterface:
"""
Return the agent in this level that gets to act in the environment
:return: Agent
"""
return list(self.agents.values())[0]
def step(self, action: Union[None, Dict[str, ActionType]]) -> EnvResponse: def step(self, action: Union[None, Dict[str, ActionType]]) -> EnvResponse:
""" """
Run a single step of following the behavioral scheme set for this environment. Run a single step of following the behavioral scheme set for this environment.
@@ -225,7 +232,7 @@ class LevelManager(EnvironmentInterface):
# step for several time steps # step for several time steps
accumulated_reward = 0 accumulated_reward = 0
acting_agent = list(self.agents.values())[0] acting_agent = self.acting_agent()
for i in range(self.steps_limit.num_steps): for i in range(self.steps_limit.num_steps):
# let the agent observe the result and decide if it wants to terminate the episode # let the agent observe the result and decide if it wants to terminate the episode
@@ -301,7 +308,7 @@ class LevelManager(EnvironmentInterface):
if self.reset_required: if self.reset_required:
self.reset_internal_state() self.reset_internal_state()
acting_agent = list(self.agents.values())[0] acting_agent = self.acting_agent()
# for i in range(self.steps_limit.num_steps): # for i in range(self.steps_limit.num_steps):
# let the agent observe the result and decide if it wants to terminate the episode # let the agent observe the result and decide if it wants to terminate the episode