from ddpg_agent import ddpg_agent from arguments import get_args from rl_utils.seeds.seeds import set_seeds from rl_utils.env_wrapper.create_env import create_single_env from mpi4py import MPI import os if __name__ == '__main__': # set thread and mpi stuff os.environ['OMP_NUM_THREADS'] = '1' os.environ['MKL_NUM_THREADS'] = '1' os.environ['IN_MPI'] = '1' # train the network args = get_args() # build up the environment env = create_single_env(args, MPI.COMM_WORLD.Get_rank()) # set the random seeds set_seeds(args, MPI.COMM_WORLD.Get_rank()) # start traininng ddpg_trainer = ddpg_agent(env, args) ddpg_trainer.learn() # close the environment env.close()
import sys from arguments import get_args from rl_utils.env_wrapper.create_env import create_single_env from rl_utils.logger import logger, bench from rl_utils.seeds.seeds import set_seeds from dqn_agent import dqn_agent import os import numpy as np if __name__ == '__main__': # get arguments args = get_args() # start to create the environment env = create_single_env(args) # set seeds set_seeds(args) # create trainer dqn_trainer = dqn_agent(env, args) # start to learn dqn_trainer.learn() # finally - close the environment env.close()