Python baselines.ppo1.cnn_policy.CnnPolicy() Examples

The following are 11 code examples of baselines.ppo1.cnn_policy.CnnPolicy(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module baselines.ppo1.cnn_policy , or try the search function .
Example #1
Source File: run_atari.py    From lirpg with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #2
Source File: run_atari.py    From HardRLWithYoutube with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank() if seed is not None else None
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #3
Source File: run_atari.py    From rl_graph_generation with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #4
Source File: run_atari.py    From DRL_DeliveryDuel with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #5
Source File: run_atari.py    From ICML2019-TREX with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank() if seed is not None else None
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #6
Source File: run_atari.py    From ICML2019-TREX with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank() if seed is not None else None
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #7
Source File: run_atari.py    From sonic_contest with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #8
Source File: run_atari.py    From self-imitation-learning with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #9
Source File: run_atari.py    From baselines with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank() if seed is not None else None
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #10
Source File: run_atari.py    From deeprl-baselines with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)
    gym.logger.setLevel(logging.WARN)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()
Example #11
Source File: run_atari.py    From BackpropThroughTheVoidRL with MIT License 5 votes vote down vote up 
def train(env_id, num_timesteps, seed):
    from baselines.ppo1 import pposgd_simple, cnn_policy
    import baselines.common.tf_util as U
    rank = MPI.COMM_WORLD.Get_rank()
    sess = U.single_threaded_session()
    sess.__enter__()
    if rank == 0:
        logger.configure()
    else:
        logger.configure(format_strs=[])
    workerseed = seed + 10000 * MPI.COMM_WORLD.Get_rank()
    set_global_seeds(workerseed)
    env = make_atari(env_id)
    def policy_fn(name, ob_space, ac_space): #pylint: disable=W0613
        return cnn_policy.CnnPolicy(name=name, ob_space=ob_space, ac_space=ac_space)
    env = bench.Monitor(env, logger.get_dir() and
        osp.join(logger.get_dir(), str(rank)))
    env.seed(workerseed)
    gym.logger.setLevel(logging.WARN)

    env = wrap_deepmind(env)
    env.seed(workerseed)

    pposgd_simple.learn(env, policy_fn,
        max_timesteps=int(num_timesteps * 1.1),
        timesteps_per_actorbatch=256,
        clip_param=0.2, entcoeff=0.01,
        optim_epochs=4, optim_stepsize=1e-3, optim_batchsize=64,
        gamma=0.99, lam=0.95,
        schedule='linear'
    )
    env.close()