def setUp(self):
agent_list = [
agents.StaticAgent(),
agents.StaticAgent(),
agents.StaticAgent(),
agents.StaticAgent(),
# agents.DockerAgent("pommerman/simple-agent", port=12345),
]
self.env = pommerman.make("PommeTeam-v0", agent_list, game_state_file='record/json/000.json')
self.env.reset()
self.env.render()
def __init__(self, config):
self.agent_list = [
agents.StaticAgent(),
agents.StaticAgent(),
agents.StaticAgent(),
agents.StaticAgent(),
]
self.env = pommerman.make(config["env_id"], self.agent_list,
config["game_state_file"])
self.is_render = config["render"]
self.action_space = self.env.action_space
self.prev_obs = None
self.stat = None
self.reset_stat()
self.observation_space = spaces.Box(low=0, high=20, shape=(17, 11, 11))
def reset(self):
self.prev_obs = self.env.reset()
obs = {}
self.reset_stat()
for i in range(4):
if self.is_agent_alive(i):
obs[i] = featurize(self.prev_obs[i])
return obs
if __name__ == '__main__':
agent_list = [
agents.RandomAgent(),
agents.StaticAgent(),
agents.StaticAgent(),
agents.StaticAgent()
]
env = pommerman.make(
'PommeTeam-v0',
agent_list,
# '/home/lucius/working/projects/pomme_rllib/resources/one_line_state.json'
)
obs = env.reset()
while True:
features = featurize(obs[0])
for i in range(17):
print(features[i])
print()
"policies_to_train": ["policy_0"],
},
"observation_filter": "MeanStdFilter",
"use_pytorch": True
},
env=v0.RllibPomme)
# fdb733b6
checkpoint = 600
checkpoint_dir = "/home/lucius/ray_results/two_policies_vs_static_agents/PPO_RllibPomme_0_2020-06-09_23-39-347whmqdrs"
ppo_agent.restore("{}/checkpoint_{}/checkpoint-{}".format(
checkpoint_dir, checkpoint, checkpoint))
agent_list = []
for agent_id in range(4):
agent_list.append(agents.StaticAgent())
env = pommerman.make("PommeTeam-v0", agent_list=agent_list)
for i in range(1):
obs = env.reset()
done = False
while not done:
env.render()
actions = env.act(obs)
actions[0] = ppo_agent.compute_action(observation=featurize(obs[0]),
policy_id="policy_0")
actions[2] = ppo_agent.compute_action(observation=featurize(obs[2]),
policy_id="policy_0")
obs, reward, done, info = env.step(actions)
print("reward:", reward)
training_agent_filename = "TEAM_ALT_AGENT" ##File name
training_agent = agents.bilal_ccritic_bignnAgent() #Teammate1
training_agent_2 = agents.bilal_ccritic_bignnAgent() #Teammate2
num_games_per_opponents = 100 #How many games per set of opponents
shuffle = True
opponents_filename = "Sta_RndNB_RulRnd_RulRndNB_Simple_SimpleNB"
opponent_lists = 6
agentsOpp = [[0] for i in range(opponent_lists)]
agentsOpp[
0] = [ ## Lists of agents to test, for now, assuming our agents are positions 0 and 2
training_agent,
agents.StaticAgent(), training_agent_2,
agents.StaticAgent()
]
agentsOpp[1] = [
training_agent,
agents.RandomAgentNoBombs(), training_agent_2,
agents.RandomAgentNoBombs()
]
agentsOpp[2] = [
training_agent,
agents.RulesRandomAgent(), training_agent_2,
agents.RulesRandomAgent()
]
agentsOpp[3] = [
training_agent,
agents.RulesRandomAgentNoBomb(), training_agent_2,