How to use the parl.Model function in parl

#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import paddle.fluid as fluid
import parl
from parl import layers
from paddle.fluid.param_attr import ParamAttr


class AtariModel(parl.Model):
    def __init__(self, act_dim):

        self.conv1 = layers.conv2d(
            num_filters=16, filter_size=4, stride=2, padding=1, act='relu')
        self.conv2 = layers.conv2d(
            num_filters=32, filter_size=4, stride=2, padding=2, act='relu')
        self.conv3 = layers.conv2d(
            num_filters=256, filter_size=11, stride=1, padding=0, act='relu')

        self.policy_conv = layers.conv2d(
            num_filters=act_dim,
            filter_size=1,
            stride=1,
            padding=0,
            act=None,
            param_attr=ParamAttr(initializer=fluid.initializer.Normal()))

#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import numpy as np
import parl
from parl import layers
from paddle import fluid
from paddle.fluid.param_attr import ParamAttr


class MujocoModel(parl.Model):
    def __init__(self, obs_dim, act_dim, init_logvar=-1.0):
        self.policy_model = PolicyModel(obs_dim, act_dim, init_logvar)
        self.value_model = ValueModel(obs_dim, act_dim)
        self.policy_lr = self.policy_model.lr
        self.value_lr = self.value_model.lr

    def policy(self, obs):
        return self.policy_model.policy(obs)

    def policy_sample(self, obs):
        return self.policy_model.sample(obs)

    def value(self, obs):
        return self.value_model.value(obs)

# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
import paddle.fluid as fluid
from parl import layers
import numpy as np
from parl import Model
from parl import Agent
from parl.utils import get_gpu_count


class RLDispatcherModel(Model):
    def __init__(self, act_dim):
        self._act_dim = act_dim
        self._fc_1 = layers.fc(size=512, act='relu')
        self._fc_2 = layers.fc(size=256, act='relu')
        self._fc_3 = layers.fc(size=128, act='tanh')
        self._output = layers.fc(size=act_dim)

    def value(self, obs):
        self._h_1 = self._fc_1(obs)
        self._h_2 = self._fc_2(self._h_1)
        self._h_3 = self._fc_3(self._h_2)
        self._pred = self._output(self._h_3)
        return self._pred

#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import parl
from parl import layers
from paddle import fluid
from paddle.fluid.param_attr import ParamAttr


class OpenSimModel(parl.Model):
    def __init__(self, obs_dim, vel_obs_dim, act_dim, model_id=0, shared=True):
        self.actor_model = ActorModel(obs_dim, vel_obs_dim, act_dim, model_id,
                                      shared)
        self.critic_model = CriticModel(obs_dim, vel_obs_dim, act_dim,
                                        model_id, shared)

    def policy(self, obs):
        return self.actor_model.policy(obs)

    def value(self, obs, action):
        return self.critic_model.value(obs, action)

    def get_actor_params(self):
        return self.actor_model.parameters()

hid1_size = 400
        hid2_size = 300

        self.fc1 = layers.fc(size=hid1_size, act='relu')
        self.fc2 = layers.fc(size=hid2_size, act='relu')
        self.fc3 = layers.fc(size=act_dim, act='tanh')

    def policy(self, obs):
        hid1 = self.fc1(obs)
        hid2 = self.fc2(hid1)
        means = self.fc3(hid2)
        means = means
        return means


class CriticModel(parl.Model):
    def __init__(self):
        hid1_size = 400
        hid2_size = 300

        self.fc1 = layers.fc(size=hid1_size, act='relu')
        self.fc2 = layers.fc(size=hid2_size, act='relu')
        self.fc3 = layers.fc(size=1, act=None)

    def value(self, obs, act):
        hid1 = self.fc1(obs)
        concat = layers.concat([hid1, act], axis=1)
        hid2 = self.fc2(concat)
        Q = self.fc3(hid2)
        Q = layers.squeeze(Q, axes=[1])
        return Q

# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import paddle.fluid as fluid
import parl
from parl import layers


class AtariModel(parl.Model):
    def __init__(self, act_dim, algo='DQN'):
        self.act_dim = act_dim

        self.conv1 = layers.conv2d(
            num_filters=32, filter_size=5, stride=1, padding=2, act='relu')
        self.conv2 = layers.conv2d(
            num_filters=32, filter_size=5, stride=1, padding=2, act='relu')
        self.conv3 = layers.conv2d(
            num_filters=64, filter_size=4, stride=1, padding=1, act='relu')
        self.conv4 = layers.conv2d(
            num_filters=64, filter_size=3, stride=1, padding=1, act='relu')

        self.algo = algo
        if algo == 'Dueling':
            self.fc1_adv = layers.fc(size=512, act='relu')
            self.fc2_adv = layers.fc(size=act_dim)

# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import parl
import paddle.fluid as fluid
from parl import layers


class AtariModel(parl.Model):
    def __init__(self, act_dim):

        self.conv1 = layers.conv2d(
            num_filters=32, filter_size=8, stride=4, padding=1, act='relu')
        self.conv2 = layers.conv2d(
            num_filters=64, filter_size=4, stride=2, padding=2, act='relu')
        self.conv3 = layers.conv2d(
            num_filters=64, filter_size=3, stride=1, padding=0, act='relu')

        self.fc = layers.fc(size=512, act='relu')

        self.policy_fc = layers.fc(size=act_dim)
        self.value_fc = layers.fc(size=1)

    def policy(self, obs):
        """

self.actor_model = ActorModel(obs_dim, vel_obs_dim, act_dim, model_id,
                                      shared)
        self.critic_model = CriticModel(obs_dim, vel_obs_dim, act_dim,
                                        model_id, shared)

    def policy(self, obs):
        return self.actor_model.policy(obs)

    def value(self, obs, action):
        return self.critic_model.value(obs, action)

    def get_actor_params(self):
        return self.actor_model.parameters()


class ActorModel(parl.Model):
    def __init__(self, obs_dim, vel_obs_dim, act_dim, model_id, shared):
        hid0_size = 800
        hid1_size = 400
        hid2_size = 200
        vel_hid0_size = 200
        vel_hid1_size = 400

        self.obs_dim = obs_dim
        self.vel_obs_dim = vel_obs_dim

        # bottom layers
        if shared:
            scope_name = 'policy_shared'
        else:
            scope_name = 'policy_identity_{}'.format(model_id)

class MujocoModel(parl.Model):
    def __init__(self, act_dim):
        self.actor_model = ActorModel(act_dim)
        self.critic_model = CriticModel()

    def policy(self, obs):
        return self.actor_model.policy(obs)

    def value(self, obs, act):
        return self.critic_model.value(obs, act)

    def get_actor_params(self):
        return self.actor_model.parameters()


class ActorModel(parl.Model):
    def __init__(self, act_dim):
        hid1_size = 400
        hid2_size = 300

        self.fc1 = layers.fc(size=hid1_size, act='relu')
        self.fc2 = layers.fc(size=hid2_size, act='relu')
        self.fc3 = layers.fc(size=act_dim, act='tanh')

    def policy(self, obs):
        hid1 = self.fc1(obs)
        hid2 = self.fc2(hid1)
        means = self.fc3(hid2)
        means = means
        return means

self.policy_model = PolicyModel(obs_dim, act_dim, init_logvar)
        self.value_model = ValueModel(obs_dim, act_dim)
        self.policy_lr = self.policy_model.lr
        self.value_lr = self.value_model.lr

    def policy(self, obs):
        return self.policy_model.policy(obs)

    def policy_sample(self, obs):
        return self.policy_model.sample(obs)

    def value(self, obs):
        return self.value_model.value(obs)


class PolicyModel(parl.Model):
    def __init__(self, obs_dim, act_dim, init_logvar):
        self.obs_dim = obs_dim
        self.act_dim = act_dim
        hid1_size = obs_dim * 10
        hid3_size = act_dim * 10
        hid2_size = int(np.sqrt(hid1_size * hid3_size))

        self.lr = 9e-4 / np.sqrt(hid2_size)

        self.fc1 = layers.fc(size=hid1_size, act='tanh')
        self.fc2 = layers.fc(size=hid2_size, act='tanh')
        self.fc3 = layers.fc(size=hid3_size, act='tanh')
        self.fc4 = layers.fc(size=act_dim, act='tanh')

        self.logvars = layers.create_parameter(
            shape=[act_dim],