policies.py

"""
Custom policies for the Cycle-of-Learning.
"""
import gym
import tensorflow as tf
import numpy as np

from stable_baselines.common.policies import BasePolicy, nature_cnn, register_policy
from stable_baselines.ddpg.policies import FeedForwardPolicy, DDPGPolicy
from stable_baselines.ddpg.noise import NormalActionNoise, OrnsteinUhlenbeckActionNoise
from stable_baselines import DDPG
from stable_baselines import DDPG
from stable_baselines.common.vec_env import DummyVecEnv

from ddpg_col import DDPG_CoL

class FeedForwardPolicyDropout(DDPGPolicy):
    """
    Policy object that implements a DDPG-like actor critic, using a feed forward neural network.
    :param sess: (TensorFlow session) The current TensorFlow session
    :param ob_space: (Gym Space) The observation space of the environment
    :param ac_space: (Gym Space) The action space of the environment
    :param n_env: (int) The number of environments to run
    :param n_steps: (int) The number of steps to run for each environment
    :param n_batch: (int) The number of batch to run (n_envs * n_steps)
    :param reuse: (bool) If the policy is reusable or not
    :param layers: ([int]) The size of the Neural network for the policy (if None, default to [64, 64])
    :param cnn_extractor: (function (TensorFlow Tensor, ``**kwargs``): (TensorFlow Tensor)) the CNN feature extraction
    :param feature_extraction: (str) The feature extraction type ("cnn" or "mlp")
    :param layer_norm: (bool) enable layer normalisation
    :param act_fun: (tf.func) the activation function to use in the neural network.
    :param kwargs: (dict) Extra keyword arguments for the nature CNN feature extraction
    """

    def __init__(self, sess, ob_space, ac_space, n_env, n_steps, n_batch, reuse=False, layers=None,
                 cnn_extractor=nature_cnn, feature_extraction="cnn",
                 layer_norm=False, act_fun=tf.nn.relu, **kwargs):
        super(FeedForwardPolicyDropout, self).__init__(sess, ob_space, ac_space, n_env, n_steps, n_batch, reuse=reuse,
                                                scale=(feature_extraction == "cnn"))

        self._kwargs_check(feature_extraction, kwargs)
        self.layer_norm = layer_norm
        self.feature_extraction = feature_extraction
        self.cnn_kwargs = kwargs
        self.cnn_extractor = cnn_extractor
        self.reuse = reuse
        self._qvalue = None
        if layers is None:
            layers = [64, 64]
        self.layers = layers

        assert len(layers) >= 1, "Error: must have at least one hidden layer for the policy."

        self.activ = act_fun

    def make_actor(self, obs=None, reuse=False, scope="pi"):
        if obs is None:
            obs = self.processed_obs

        with tf.variable_scope(scope, reuse=reuse):
            # setup dropout and mechanisms to enable it during test time only
            self.prob_dropout_ph = tf.placeholder_with_default(0.0, shape=(), name='prob_dropout_ph')

            if self.feature_extraction == "cnn":
                pi_h = self.cnn_extractor(obs, **self.cnn_kwargs)
            else:
                pi_h = tf.layers.flatten(obs)
            for i, layer_size in enumerate(self.layers):
                pi_h = tf.nn.dropout(pi_h, rate=self.prob_dropout_ph)
                pi_h = tf.layers.dense(pi_h, layer_size, name='fc' + str(i))
                if self.layer_norm:
                    pi_h = tf.contrib.layers.layer_norm(pi_h, center=True, scale=True)
                pi_h = self.activ(pi_h)
                pi_h = tf.nn.dropout(pi_h, rate=self.prob_dropout_ph)
            self.policy = tf.nn.tanh(tf.layers.dense(pi_h, self.ac_space.shape[0], name=scope,
                                                     kernel_initializer=tf.random_uniform_initializer(minval=-3e-3,
                                                                                                      maxval=3e-3)))
        return self.policy

    def make_critic(self, obs=None, action=None, reuse=False, scope="qf"):
        if obs is None:
            obs = self.processed_obs
        if action is None:
            action = self.action_ph

        with tf.variable_scope(scope, reuse=reuse):
            if self.feature_extraction == "cnn":
                qf_h = self.cnn_extractor(obs, **self.cnn_kwargs)
            else:
                qf_h = tf.layers.flatten(obs)
            for i, layer_size in enumerate(self.layers):
                qf_h = tf.layers.dense(qf_h, layer_size, name='fc' + str(i))
                if self.layer_norm:
                    qf_h = tf.contrib.layers.layer_norm(qf_h, center=True, scale=True)
                qf_h = self.activ(qf_h)
                if i == 0:
                    qf_h = tf.concat([qf_h, action], axis=-1)

            # the name attribute is used in pop-art normalization
            qvalue_fn = tf.layers.dense(qf_h, 1, name='qf_output',
                                        kernel_initializer=tf.random_uniform_initializer(minval=-3e-3,
                                                                                         maxval=3e-3))

            self.qvalue_fn = qvalue_fn
            self._qvalue = qvalue_fn[:, 0]
        return self.qvalue_fn

    def step(self, obs, state=None, mask=None):
        return self.sess.run(self.policy, {self.obs_ph: obs})

    def proba_step(self, obs, state=None, mask=None):
        return self.sess.run(self.policy, {self.obs_ph: obs})

    def value(self, obs, action, state=None, mask=None):
        return self.sess.run(self._qvalue, {self.obs_ph: obs, self.action_ph: action})


class MlpPolicyDropout(FeedForwardPolicyDropout):
    """
    Policy object that implements actor critic, using a MLP (2 layers of 64)
    :param sess: (TensorFlow session) The current TensorFlow session
    :param ob_space: (Gym Space) The observation space of the environment
    :param ac_space: (Gym Space) The action space of the environment
    :param n_env: (int) The number of environments to run
    :param n_steps: (int) The number of steps to run for each environment
    :param n_batch: (int) The number of batch to run (n_envs * n_steps)
    :param reuse: (bool) If the policy is reusable or not
    :param _kwargs: (dict) Extra keyword arguments for the nature CNN feature extraction
    """

    def __init__(self, sess, ob_space, ac_space, n_env, n_steps, n_batch, reuse=False, **_kwargs):
        super(MlpPolicyDropout, self).__init__(sess, ob_space, ac_space, n_env, n_steps, n_batch, reuse,
                                        feature_extraction="mlp", **_kwargs)

def dump(obj):
    """ Prints attributes of any object."""
    for attr in dir(obj):
        print("obj.%s = %r" % (attr, getattr(obj, attr)))

if __name__ == "__main__":
    # create and wrap the environment
    env = DummyVecEnv([lambda: gym.make('LunarLanderContinuous-v2')])

    # the noise objects for DDPG
    n_actions = env.action_space.shape[-1]
    param_noise = None
    action_noise = OrnsteinUhlenbeckActionNoise(mean=np.zeros(n_actions), sigma=float(0.5) * np.ones(n_actions))

    # create model using custom policy
    model = DDPG(MlpPolicyDropout, env, verbose=2, param_noise=param_noise, action_noise=action_noise)
    model.learn(total_timesteps=1000)

    # test dropout
    print('****** Testing dropout ******')
    obs = np.random.rand(8)
    for i in range(20):
        action, qval = model._policy(obs, apply_noise=False, compute_q=True)
        print(f'action: {action} | qval: {qval[0]}')