from rlcard3.games.mocsar.util_examples import init_environment, init_vars
# Config
conf = Config('environ.properties')
# Environemtn
env, eval_env = init_environment(conf=conf,
env_id='mocsar-cfg',
config={'multi_agent_mode': True})
# parameter variables
evaluate_num, evaluate_every, memory_init_size, train_every, episode_num = init_vars(
conf=conf)
# The paths for saving the logs and learning curves
log_dir = './experiments/mocsar_dqn_ra_pytorch_result/'
# Set a global seed
set_global_seed(0)
agent = DQNAgent(scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
train_every=train_every,
state_shape=env.state_shape,
mlp_layers=[512, 512],
device=torch.device('cuda'))
random_agent = RandomAgent(action_num=eval_env.action_num)
# Other agents
env.model.create_agents({"mocsar_min": 4})
env_agent_list = [env.model.rule_agents[i] for i in range(1, 4)]
env_agent_list.insert(0, agent)
# Set the number of process
process_num = 8
# Set episode_num
episode_num = 10000
# Assign tasks
per_tasks = assign_task(episode_num, process_num)
# Set game and make environment
game = 'doudizhu'
env = rlcard3.make(game)
# Set global seed
set_global_seed(1)
# Set up agents
agent_num = env.player_num
env.set_agents(
[RandomAgent(action_num=env.action_num) for _ in range(agent_num)])
# Set a global list to reserve trajectories
manager = multiprocessing.Manager()
trajectories_set = manager.list()
# Generate Processes
processes = []
for p in range(process_num):
process = multiprocessing.Process(target=env.run_multi,
args=(per_tasks[p],
def test_set_global_seed(self):
set_global_seed(0)
self.assertEqual(np.random.get_state()[1][0], 0)
""" A toy example of playing Mocsar with random agents
"""
import tensorflow as tf
from rlcard3.agents.dqn_agent import DQNAgent
import rlcard3
from rlcard3.utils.utils import set_global_seed
from rlcard3.model_agents.registration import get_agents
# Make environment
env = rlcard3.make('mocsar')
episode_num = 2
# Set a global seed
set_global_seed(seed=0)
# Set up agents
agents = {"mocsar_random": 3, "mocsar_min": 1}
# Create DQN agent
# Set a global seed
set_global_seed(0)
# Load pretrained model
graph = tf.Graph()
sess = tf.Session(graph=graph)
with graph.as_default():
agent = DQNAgent(sess,
scope='dqn',
action_num=env.action_num,
replay_memory_init_size=memory_init_size,
""" A toy example of playing Mocsar with random agents
"""
import rlcard3
from rlcard3.utils.utils import set_global_seed
from rlcard3.model_agents.registration import get_agents
# Make environment
env = rlcard3.make('mocsar')
episode_num = 2
# Set a global seed
set_global_seed(seed=0) # TODO ez még nem fut...
# Set up agents
agents = {"mocsar_random": 2, "mocsar_min": 2}
env.set_agents(get_agents(agents=agents, nr_players=4))
for episode in range(episode_num):
# Generate data from the environment
trajectories, _ = env.run(is_training=False)
# Print out the trajectories
print('\nEpisode {}'.format(episode))
for ts in trajectories[0]:
print('State: {}, Action: {}, Reward: {}, Next State: {}, Done: {}'.
format(ts[0], ts[1], ts[2], ts[3], ts[4]))
