garage.tf.baselines
¶
Baseline estimators for TensorFlow-based algorithms.
- class ContinuousMLPBaseline(env_spec, num_seq_inputs=1, name='ContinuousMLPBaseline', hidden_sizes=(32, 32), hidden_nonlinearity=tf.nn.tanh, hidden_w_init=tf.initializers.glorot_uniform(seed=deterministic.get_tf_seed_stream()), hidden_b_init=tf.zeros_initializer(), output_nonlinearity=None, output_w_init=tf.initializers.glorot_uniform(seed=deterministic.get_tf_seed_stream()), output_b_init=tf.zeros_initializer(), optimizer=None, optimizer_args=None, normalize_inputs=True)¶
Bases:
garage.tf.models.NormalizedInputMLPModel
,garage.np.baselines.Baseline
A value function using a MLP network.
It fits the input data by performing linear regression to the outputs.
- Parameters
env_spec (garage.envs.env_spec.EnvSpec) – Environment specification.
num_seq_inputs (float) – Number of sequence per input. By default it is 1.0, which means only one single sequence.
name (str) – Name of baseline.
hidden_sizes (list[int]) – Output dimension of dense layer(s) for the MLP for mean. For example, (32, 32) means the MLP consists of two hidden layers, each with 32 hidden units.
hidden_nonlinearity (Callable) – Activation function for intermediate dense layer(s). It should return a tf.Tensor. Set it to None to maintain a linear activation.
hidden_w_init (Callable) – Initializer function for the weight of intermediate dense layer(s). The function should return a tf.Tensor.
hidden_b_init (Callable) – Initializer function for the bias of intermediate dense layer(s). The function should return a tf.Tensor.
output_nonlinearity (Callable) – Activation function for output dense layer. It should return a tf.Tensor. Set it to None to maintain a linear activation.
output_w_init (Callable) – Initializer function for the weight of output dense layer(s). The function should return a tf.Tensor.
output_b_init (Callable) – Initializer function for the bias of output dense layer(s). The function should return a tf.Tensor.
optimizer (garage.tf.Optimizer) – Optimizer for minimizing the negative log-likelihood.
optimizer_args (dict) – Arguments for the optimizer. Default is None, which means no arguments.
normalize_inputs (bool) – Bool for normalizing inputs or not.
- property env_spec¶
Policy environment specification.
- Returns
Environment specification.
- Return type
- property parameters¶
Parameters of the model.
- Returns
Parameters
- Return type
np.ndarray
- property name¶
Name (str) of the model.
This is also the variable scope of the model.
- Returns
Name of the model.
- Return type
- property input¶
Default input of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the input of the network.
- Returns
Default input of the model.
- Return type
tf.Tensor
- property output¶
Default output of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the output of the network.
- Returns
Default output of the model.
- Return type
tf.Tensor
- property inputs¶
Default inputs of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the inputs of the network.
- Returns
Default inputs of the model.
- Return type
list[tf.Tensor]
- property outputs¶
Default outputs of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the outputs of the network.
- Returns
Default outputs of the model.
- Return type
list[tf.Tensor]
- property state_info_specs¶
State info specification.
- Returns
- keys and shapes for the information related to the
module’s state when taking an action.
- Return type
List[str]
- property state_info_keys¶
State info keys.
- Returns
- keys for the information related to the module’s state
when taking an input.
- Return type
List[str]
- predict(paths)¶
Predict value based on paths.
- Parameters
paths (dict[numpy.ndarray]) – Sample paths.
- Returns
Predicted value.
- Return type
numpy.ndarray
- network_output_spec()¶
Network output spec.
- build(*inputs, name=None)¶
Build a Network with the given input(s).
* Do not call tf.global_variable_initializers() after building a model as it will reassign random weights to the model. The parameters inside a model will be initialized when calling build(). *
It uses the same, fixed variable scope for all Networks, to ensure parameter sharing. Different Networks must have an unique name.
- Parameters
- Raises
ValueError – When a Network with the same name is already built.
- Returns
- Output tensors of the model with the given
inputs.
- Return type
list[tf.Tensor]
- network_input_spec()¶
Network input spec.
- reset(do_resets=None)¶
Reset the module.
This is effective only to recurrent modules. do_resets is effective only to vectoried modules.
For a vectorized modules, do_resets is an array of boolean indicating which internal states to be reset. The length of do_resets should be equal to the length of inputs.
- Parameters
do_resets (numpy.ndarray) – Bool array indicating which states to be reset.
- terminate()¶
Clean up operation.
- get_trainable_vars()¶
Get trainable variables.
- Returns
- A list of trainable variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_global_vars()¶
Get global variables.
- Returns
- A list of global variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_regularizable_vars()¶
Get all network weight variables in the current scope.
- Returns
- A list of network weight variables in the
current variable scope.
- Return type
List[tf.Variable]
- get_params()¶
Get the trainable variables.
- Returns
- A list of trainable variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_param_shapes()¶
Get parameter shapes.
- Returns
A list of variable shapes.
- Return type
List[tuple]
- get_param_values()¶
Get param values.
- Returns
- Values of the parameters evaluated in
the current session
- Return type
np.ndarray
- set_param_values(param_values)¶
Set param values.
- Parameters
param_values (np.ndarray) – A numpy array of parameter values.
- flat_to_params(flattened_params)¶
Unflatten tensors according to their respective shapes.
- Parameters
flattened_params (np.ndarray) – A numpy array of flattened params.
- Returns
- A list of parameters reshaped to the
shapes specified.
- Return type
List[np.ndarray]
- class GaussianCNNBaseline(env_spec, filters, strides, padding, hidden_sizes, hidden_nonlinearity=tf.nn.tanh, hidden_w_init=tf.initializers.glorot_uniform(seed=deterministic.get_tf_seed_stream()), hidden_b_init=tf.zeros_initializer(), output_nonlinearity=None, output_w_init=tf.initializers.glorot_uniform(seed=deterministic.get_tf_seed_stream()), output_b_init=tf.zeros_initializer(), name='GaussianCNNBaseline', learn_std=True, init_std=1.0, adaptive_std=False, std_share_network=False, std_filters=(), std_strides=(), std_padding='SAME', std_hidden_sizes=(), std_hidden_nonlinearity=None, std_output_nonlinearity=None, layer_normalization=False, normalize_inputs=True, normalize_outputs=True, subsample_factor=1.0, optimizer=None, optimizer_args=None, use_trust_region=True, max_kl_step=0.01)¶
Bases:
garage.tf.baselines.gaussian_cnn_baseline_model.GaussianCNNBaselineModel
,garage.np.baselines.baseline.Baseline
Fits a Gaussian distribution to the outputs of a CNN.
- Parameters
env_spec (garage.envs.env_spec.EnvSpec) – Environment specification.
filters (Tuple[Tuple[int, Tuple[int, int]], ...]) – Number and dimension of filters. For example, ((3, (3, 5)), (32, (3, 3))) means there are two convolutional layers. The filter for the first layer have 3 channels and its shape is (3 x 5), while the filter for the second layer have 32 channels and its shape is (3 x 3).
strides (tuple[int]) – The stride of the sliding window. For example, (1, 2) means there are two convolutional layers. The stride of the filter for first layer is 1 and that of the second layer is 2.
padding (str) – The type of padding algorithm to use, either ‘SAME’ or ‘VALID’.
name (str) – Model name, also the variable scope.
hidden_sizes (list[int]) – Output dimension of dense layer(s) for the Convolutional model for mean. For example, (32, 32) means the network consists of two dense layers, each with 32 hidden units.
hidden_nonlinearity (Callable) – Activation function for intermediate dense layer(s). It should return a tf.Tensor. Set it to None to maintain a linear activation.
hidden_w_init (Callable) – Initializer function for the weight of intermediate dense layer(s). The function should return a tf.Tensor.
hidden_b_init (Callable) – Initializer function for the bias of intermediate dense layer(s). The function should return a tf.Tensor.
output_nonlinearity (Callable) – Activation function for output dense layer. It should return a tf.Tensor. Set it to None to maintain a linear activation.
output_w_init (Callable) – Initializer function for the weight of output dense layer(s). The function should return a tf.Tensor.
output_b_init (Callable) – Initializer function for the bias of output dense layer(s). The function should return a tf.Tensor.
name – Name of this model (also used as its scope).
learn_std (bool) – Whether to train the standard deviation parameter of the Gaussian distribution.
init_std (float) – Initial standard deviation for the Gaussian distribution.
adaptive_std (bool) – Whether to use a neural network to learn the standard deviation of the Gaussian distribution. Unless True, the standard deviation is learned as a parameter which is not conditioned on the inputs.
std_share_network (bool) – Boolean for whether the mean and standard deviation models share a CNN network. If True, each is a head from a single body network. Otherwise, the parameters are estimated using the outputs of two indepedent networks.
std_filters (Tuple[Tuple[int, Tuple[int, int]], ...]) – Number and dimension of filters. For example, ((3, (3, 5)), (32, (3, 3))) means there are two convolutional layers. The filter for the first layer have 3 channels and its shape is (3 x 5), while the filter for the second layer have 32 channels and its shape is (3 x 3).
std_strides (tuple[int]) – The stride of the sliding window. For example, (1, 2) means there are two convolutional layers. The stride of the filter for first layer is 1 and that of the second layer is 2.
std_padding (str) – The type of padding algorithm to use in std network, either ‘SAME’ or ‘VALID’.
std_hidden_sizes (list[int]) – Output dimension of dense layer(s) for the Conv for std. For example, (32, 32) means the Conv consists of two hidden layers, each with 32 hidden units.
std_hidden_nonlinearity (callable) – Nonlinearity for each hidden layer in the std network.
std_output_nonlinearity (Callable) – Activation function for output dense layer in the std network. It should return a tf.Tensor. Set it to None to maintain a linear activation.
layer_normalization (bool) – Bool for using layer normalization or not.
normalize_inputs (bool) – Bool for normalizing inputs or not.
normalize_outputs (bool) – Bool for normalizing outputs or not.
subsample_factor (float) – The factor to subsample the data. By default it is 1.0, which means using all the data.
optimizer (garage.tf.Optimizer) – Optimizer used for fitting the model.
optimizer_args (dict) – Arguments for the optimizer. Default is None, which means no arguments.
use_trust_region (bool) – Whether to use a KL-divergence constraint.
max_kl_step (float) – KL divergence constraint for each iteration, if use_trust_region is active.
- property env_spec¶
Policy environment specification.
- Returns
Environment specification.
- Return type
- property parameters¶
Parameters of the model.
- Returns
Parameters
- Return type
np.ndarray
- property name¶
Name (str) of the model.
This is also the variable scope of the model.
- Returns
Name of the model.
- Return type
- property input¶
Default input of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the input of the network.
- Returns
Default input of the model.
- Return type
tf.Tensor
- property output¶
Default output of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the output of the network.
- Returns
Default output of the model.
- Return type
tf.Tensor
- property inputs¶
Default inputs of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the inputs of the network.
- Returns
Default inputs of the model.
- Return type
list[tf.Tensor]
- property outputs¶
Default outputs of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the outputs of the network.
- Returns
Default outputs of the model.
- Return type
list[tf.Tensor]
- property state_info_specs¶
State info specification.
- Returns
- keys and shapes for the information related to the
module’s state when taking an action.
- Return type
List[str]
- property state_info_keys¶
State info keys.
- Returns
- keys for the information related to the module’s state
when taking an input.
- Return type
List[str]
- predict(paths)¶
Predict ys based on input xs.
- Parameters
paths (dict[numpy.ndarray]) – Sample paths.
- Returns
The predicted ys.
- Return type
numpy.ndarray
- clone_model(name)¶
Return a clone of the GaussianCNNBaselineModel.
It copies the configuration of the primitive and also the parameters.
- Parameters
name (str) – Name of the newly created model. It has to be different from source policy if cloned under the same computational graph.
- Returns
Newly cloned model.
- Return type
garage.tf.baselines.GaussianCNNBaselineModel
- network_output_spec()¶
Network output spec.
- build(*inputs, name=None)¶
Build a Network with the given input(s).
* Do not call tf.global_variable_initializers() after building a model as it will reassign random weights to the model. The parameters inside a model will be initialized when calling build(). *
It uses the same, fixed variable scope for all Networks, to ensure parameter sharing. Different Networks must have an unique name.
- Parameters
- Raises
ValueError – When a Network with the same name is already built.
- Returns
- Output tensors of the model with the given
inputs.
- Return type
list[tf.Tensor]
- network_input_spec()¶
Network input spec.
- reset(do_resets=None)¶
Reset the module.
This is effective only to recurrent modules. do_resets is effective only to vectoried modules.
For a vectorized modules, do_resets is an array of boolean indicating which internal states to be reset. The length of do_resets should be equal to the length of inputs.
- Parameters
do_resets (numpy.ndarray) – Bool array indicating which states to be reset.
- terminate()¶
Clean up operation.
- get_trainable_vars()¶
Get trainable variables.
- Returns
- A list of trainable variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_global_vars()¶
Get global variables.
- Returns
- A list of global variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_regularizable_vars()¶
Get all network weight variables in the current scope.
- Returns
- A list of network weight variables in the
current variable scope.
- Return type
List[tf.Variable]
- get_params()¶
Get the trainable variables.
- Returns
- A list of trainable variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_param_shapes()¶
Get parameter shapes.
- Returns
A list of variable shapes.
- Return type
List[tuple]
- get_param_values()¶
Get param values.
- Returns
- Values of the parameters evaluated in
the current session
- Return type
np.ndarray
- set_param_values(param_values)¶
Set param values.
- Parameters
param_values (np.ndarray) – A numpy array of parameter values.
- flat_to_params(flattened_params)¶
Unflatten tensors according to their respective shapes.
- Parameters
flattened_params (np.ndarray) – A numpy array of flattened params.
- Returns
- A list of parameters reshaped to the
shapes specified.
- Return type
List[np.ndarray]
- class GaussianMLPBaseline(env_spec, num_seq_inputs=1, name='GaussianMLPBaseline', hidden_sizes=(32, 32), hidden_nonlinearity=tf.nn.tanh, hidden_w_init=tf.initializers.glorot_uniform(seed=deterministic.get_tf_seed_stream()), hidden_b_init=tf.zeros_initializer(), output_nonlinearity=None, output_w_init=tf.initializers.glorot_uniform(seed=deterministic.get_tf_seed_stream()), output_b_init=tf.zeros_initializer(), optimizer=None, optimizer_args=None, use_trust_region=True, max_kl_step=0.01, learn_std=True, init_std=1.0, adaptive_std=False, std_share_network=False, std_hidden_sizes=(32, 32), std_nonlinearity=None, layer_normalization=False, normalize_inputs=True, normalize_outputs=True, subsample_factor=1.0)¶
Bases:
garage.tf.baselines.gaussian_mlp_baseline_model.GaussianMLPBaselineModel
,garage.np.baselines.Baseline
Gaussian MLP Baseline with Model.
It fits the input data to a gaussian distribution estimated by a MLP.
- Parameters
env_spec (garage.envs.env_spec.EnvSpec) – Environment specification.
subsample_factor (float) – The factor to subsample the data. By default it is 1.0, which means using all the data.
num_seq_inputs (float) – Number of sequence per input. By default it is 1.0, which means only one single sequence.
name (str) – Name of baseline.
hidden_sizes (list[int]) – Output dimension of dense layer(s) for the MLP for mean. For example, (32, 32) means the MLP consists of two hidden layers, each with 32 hidden units.
hidden_nonlinearity (Callable) – Activation function for intermediate dense layer(s). It should return a tf.Tensor. Set it to None to maintain a linear activation.
hidden_w_init (Callable) – Initializer function for the weight of intermediate dense layer(s). The function should return a tf.Tensor.
hidden_b_init (Callable) – Initializer function for the bias of intermediate dense layer(s). The function should return a tf.Tensor.
output_nonlinearity (Callable) – Activation function for output dense layer. It should return a tf.Tensor. Set it to None to maintain a linear activation.
output_w_init (Callable) – Initializer function for the weight of output dense layer(s). The function should return a tf.Tensor.
output_b_init (Callable) – Initializer function for the bias of output dense layer(s). The function should return a tf.Tensor.
optimizer (garage.tf.Optimizer) – Optimizer for minimizing the negative log-likelihood.
optimizer_args (dict) – Arguments for the optimizer. Default is None, which means no arguments.
use_trust_region (bool) – Whether to use trust region constraint.
max_kl_step (float) – KL divergence constraint for each iteration.
learn_std (bool) – Is std trainable.
init_std (float) – Initial value for std.
adaptive_std (bool) – Is std a neural network. If False, it will be a parameter.
std_share_network (bool) – Boolean for whether mean and std share the same network.
std_hidden_sizes (list[int]) – Output dimension of dense layer(s) for the MLP for std. For example, (32, 32) means the MLP consists of two hidden layers, each with 32 hidden units.
std_nonlinearity (Callable) – Nonlinearity for each hidden layer in the std network.
layer_normalization (bool) – Bool for using layer normalization or not.
normalize_inputs (bool) – Bool for normalizing inputs or not.
normalize_outputs (bool) – Bool for normalizing outputs or not.
subsample_factor – The factor to subsample the data. By default it is 1.0, which means using all the data.
- property env_spec¶
Policy environment specification.
- Returns
Environment specification.
- Return type
- property parameters¶
Parameters of the model.
- Returns
Parameters
- Return type
np.ndarray
- property name¶
Name (str) of the model.
This is also the variable scope of the model.
- Returns
Name of the model.
- Return type
- property input¶
Default input of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the input of the network.
- Returns
Default input of the model.
- Return type
tf.Tensor
- property output¶
Default output of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the output of the network.
- Returns
Default output of the model.
- Return type
tf.Tensor
- property inputs¶
Default inputs of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the inputs of the network.
- Returns
Default inputs of the model.
- Return type
list[tf.Tensor]
- property outputs¶
Default outputs of the model.
When the model is built the first time, by default it creates the ‘default’ network. This property creates a reference to the outputs of the network.
- Returns
Default outputs of the model.
- Return type
list[tf.Tensor]
- property state_info_specs¶
State info specification.
- Returns
- keys and shapes for the information related to the
module’s state when taking an action.
- Return type
List[str]
- property state_info_keys¶
State info keys.
- Returns
- keys for the information related to the module’s state
when taking an input.
- Return type
List[str]
- predict(paths)¶
Predict value based on paths.
- clone_model(name)¶
Return a clone of the GaussianMLPBaselineModel.
It copies the configuration of the primitive and also the parameters.
- Parameters
name (str) – Name of the newly created model. It has to be different from source policy if cloned under the same computational graph.
- Returns
Newly cloned model.
- Return type
garage.tf.baselines.GaussianMLPBaselineModel
- network_output_spec()¶
Network output spec.
- build(*inputs, name=None)¶
Build a Network with the given input(s).
* Do not call tf.global_variable_initializers() after building a model as it will reassign random weights to the model. The parameters inside a model will be initialized when calling build(). *
It uses the same, fixed variable scope for all Networks, to ensure parameter sharing. Different Networks must have an unique name.
- Parameters
- Raises
ValueError – When a Network with the same name is already built.
- Returns
- Output tensors of the model with the given
inputs.
- Return type
list[tf.Tensor]
- network_input_spec()¶
Network input spec.
- reset(do_resets=None)¶
Reset the module.
This is effective only to recurrent modules. do_resets is effective only to vectoried modules.
For a vectorized modules, do_resets is an array of boolean indicating which internal states to be reset. The length of do_resets should be equal to the length of inputs.
- Parameters
do_resets (numpy.ndarray) – Bool array indicating which states to be reset.
- terminate()¶
Clean up operation.
- get_trainable_vars()¶
Get trainable variables.
- Returns
- A list of trainable variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_global_vars()¶
Get global variables.
- Returns
- A list of global variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_regularizable_vars()¶
Get all network weight variables in the current scope.
- Returns
- A list of network weight variables in the
current variable scope.
- Return type
List[tf.Variable]
- get_params()¶
Get the trainable variables.
- Returns
- A list of trainable variables in the current
variable scope.
- Return type
List[tf.Variable]
- get_param_shapes()¶
Get parameter shapes.
- Returns
A list of variable shapes.
- Return type
List[tuple]
- get_param_values()¶
Get param values.
- Returns
- Values of the parameters evaluated in
the current session
- Return type
np.ndarray
- set_param_values(param_values)¶
Set param values.
- Parameters
param_values (np.ndarray) – A numpy array of parameter values.
- flat_to_params(flattened_params)¶
Unflatten tensors according to their respective shapes.
- Parameters
flattened_params (np.ndarray) – A numpy array of flattened params.
- Returns
- A list of parameters reshaped to the
shapes specified.
- Return type
List[np.ndarray]