def default_policy(self, node: MPMCTSTreeNode, env: Env):
if env.is_over():
return env.get_payoffs()
player_id = env.get_player_id()
#print(player_id)
action = random.sample(node.legal_actions[player_id], 1)[0]
while not env.is_over():
# step forward
next_state, next_player_id = env.step(action, False)
if not env.is_over():
action = random.sample(next_state['legal_actions'], 1)[0]
# game over
return env.get_payoffs()
def default_policy(self, node: MPMCTSTreeNode, env: Env):
if env.is_over():
return env.get_payoffs()
player_id = env.get_player_id()
#print(player_id)
state = env.get_state(player_id)
action, _ = self.drqn_agents[player_id].eval_step(state)
while not env.is_over():
#print(action)
# step forward
next_state, next_player_id = env.step(action, False)
if not env.is_over():
#action = random.sample(next_state['legal_actions'],1)[0]
action, _ = self.drqn_agents[next_player_id].eval_step(
next_state)
# game over
return env.get_payoffs()