class SimpleGaussianMLPRegressor(StochasticRegressor2):
"""Simple GaussianMLPRegressor for testing."""
def __init__(self, input_shape, output_dim, name, *args, **kwargs):
super().__init__(input_shape, output_dim, name)
self.model = SimpleGaussianMLPModel(output_dim=self._output_dim)
self._initialize()
def _initialize(self):
input_ph = tf.placeholder(tf.float32,
shape=(None, ) + self._input_shape)
with tf.variable_scope(self._name) as self._variable_scope:
self.model.build(input_ph)
self.ys = None
def fit(self, xs, ys):
self.ys = ys
def predict(self, xs):
if self.ys is None:
mean = tf.get_default_session().run(
self.model.networks['default'].mean,
feed_dict={self.model.networks['default'].input: xs})
self.ys = np.full((len(xs), 1), mean)
return self.ys
def get_params_internal(self, *args, **kwargs):
return self._variable_scope.trainable_variables()
def __setstate__(self, state):
"""Object.__setstate__."""
super().__setstate__(state)
self._initialize()
class SimpleGaussianMLPRegressor(StochasticRegressor):
"""Simple GaussianMLPRegressor for testing.
Args:
input_shape (tuple[int]): Input shape of the training data.
output_dim (int): Output dimension of the model.
name (str): Model name, also the variable scope.
args (list): Unused positionl arguments.
kwargs (dict): Unused keyword arguments.
"""
def __init__(self, input_shape, output_dim, name, *args, **kwargs):
super().__init__(input_shape, output_dim, name)
del args, kwargs
self.model = SimpleGaussianMLPModel(output_dim=self._output_dim)
self._ys = None
self._network = None
self._initialize()
@property
def recurrent(self):
"""bool: If this module has a hidden state."""
return False
@property
def vectorized(self):
"""bool: If this module supports vectorization input."""
return True
@property
def distribution(self):
"""garage.tf.distributions.DiagonalGaussian: Distribution."""
return self._network.dist
def dist_info_sym(self, input_var, state_info_vars=None, name='default'):
"""Create a symbolic graph of the distribution parameters.
Args:
input_var (tf.Tensor): tf.Tensor of the input data.
state_info_vars (dict): a dictionary whose values should contain
information about the state of the policy at the time it
received the input.
name (str): Name of the new graph.
Return:
dict[tf.Tensor]: Outputs of the symbolic distribution parameter
graph.
"""
del state_info_vars
with tf.compat.v1.variable_scope(self._variable_scope):
network = self.model.build(input_var, name=name)
means_var = network.means
log_stds_var = network.log_stds
return dict(mean=means_var, log_std=log_stds_var)
def _initialize(self):
"""Initialize graph."""
input_ph = tf.compat.v1.placeholder(tf.float32,
shape=(None, ) + self._input_shape)
with tf.compat.v1.variable_scope(self._name) as vs:
self._variable_scope = vs
self._network = self.model.build(input_ph)
def fit(self, xs, ys):
"""Fit with input data xs and label ys.
Args:
xs (numpy.ndarray): Input data.
ys (numpy.ndarray): Label of input data.
"""
self._ys = ys
def predict(self, xs):
"""Predict ys based on input xs.
Args:
xs (numpy.ndarray): Input data.
Return:
np.ndarray: The predicted ys.
"""
if self._ys is None:
mean = tf.compat.v1.get_default_session().run(
self._network.mean, feed_dict={self._network.input: xs})
self._ys = np.full((len(xs), 1), mean)
return self._ys
def get_params_internal(self):
"""Get the params, which are the trainable variables.
Returns:
List[tf.Variable]: A list of trainable variables in the current
variable scope.
"""
return self._variable_scope.trainable_variables()
def __getstate__(self):
"""Object.__getstate__.
Returns:
dict: the state to be pickled for the instance.
"""
new_dict = super().__getstate__()
del new_dict['_network']
return new_dict
def __setstate__(self, state):
"""Object.__setstate__.
Args:
state (dict): Unpickled state.
"""
super().__setstate__(state)
self._initialize()
class SimpleGaussianMLPRegressor(StochasticRegressor):
"""Simple GaussianMLPRegressor for testing.
Args:
input_shape (tuple[int]): Input shape of the training data.
output_dim (int): Output dimension of the model.
name (str): Model name, also the variable scope.
args (list): Unused positionl arguments.
kwargs (dict): Unused keyword arguments.
"""
# pylint: disable=unused-argument
def __init__(self, input_shape, output_dim, name, *args, **kwargs):
super().__init__(input_shape, output_dim, name)
self.model = SimpleGaussianMLPModel(output_dim=self._output_dim)
self._ys = None
self._initialize()
def _initialize(self):
"""Initialize graph."""
input_ph = tf.compat.v1.placeholder(tf.float32,
shape=(None, ) + self._input_shape)
with tf.compat.v1.variable_scope(self._name) as vs:
self._variable_scope = vs
self.model.build(input_ph)
def fit(self, xs, ys):
"""Fit with input data xs and label ys.
Args:
xs (numpy.ndarray): Input data.
ys (numpy.ndarray): Label of input data.
"""
self._ys = ys
def predict(self, xs):
"""Predict ys based on input xs.
Args:
xs (numpy.ndarray): Input data.
Return:
np.ndarray: The predicted ys.
"""
if self._ys is None:
mean = tf.compat.v1.get_default_session().run(
self.model.networks['default'].mean,
feed_dict={self.model.networks['default'].input: xs})
self._ys = np.full((len(xs), 1), mean)
return self._ys
def get_params_internal(self):
"""Get the params, which are the trainable variables.
Returns:
List[tf.Variable]: A list of trainable variables in the current
variable scope.
"""
return self._variable_scope.trainable_variables()
def __setstate__(self, state):
"""Object.__setstate__.
Args:
state (dict): Unpickled state.
"""
super().__setstate__(state)
self._initialize()