garage.tf.q_functions.continuous_cnn_q_function module

Continuous CNN QFunction with CNN-MLP structure.

class ContinuousCNNQFunction(env_spec, filters, strides, hidden_sizes=(256, ), action_merge_layer=-2, name=None, padding='SAME', max_pooling=False, pool_strides=(2, 2), pool_shapes=(2, 2), cnn_hidden_nonlinearity=<function relu>, hidden_nonlinearity=<function relu>, hidden_w_init=<tensorflow.python.ops.init_ops_v2.GlorotUniform object>, hidden_b_init=<tensorflow.python.ops.init_ops_v2.Zeros object>, output_nonlinearity=None, output_w_init=<tensorflow.python.ops.init_ops_v2.GlorotUniform object>, output_b_init=<tensorflow.python.ops.init_ops_v2.Zeros object>, layer_normalization=False)[source]

Bases: garage.tf.q_functions.q_function.QFunction

Q function based on a CNN-MLP structure for continuous action space.

This class implements a Q value network to predict Q based on the input state and action. It uses an CNN and a MLP to fit the function of Q(s, a).

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.
  • hidden_sizes (tuple[int]) – Output dimension of dense layer(s). For example, (32, 32) means the MLP of this q-function consists of two hidden layers, each with 32 hidden units.
  • action_merge_layer (int) – The index of layers at which to concatenate action inputs with the network. The indexing works like standard python list indexing. Index of 0 refers to the input layer (observation input) while an index of -1 points to the last hidden layer. Default parameter points to second layer from the end.
  • name (str) – Variable scope of the cnn.
  • padding (str) – The type of padding algorithm to use, either ‘SAME’ or ‘VALID’.
  • max_pooling (bool) – Boolean for using max pooling layer or not.
  • pool_shapes (tuple[int]) – Dimension of the pooling layer(s). For example, (2, 2) means that all the pooling layers have shape (2, 2).
  • pool_strides (tuple[int]) – The strides of the pooling layer(s). For example, (2, 2) means that all the pooling layers have strides (2, 2).
  • cnn_hidden_nonlinearity (callable) – Activation function for intermediate dense layer(s) in the CNN. It should return a tf.Tensor. Set it to None to maintain a linear activation.
  • hidden_nonlinearity (callable) – Activation function for intermediate dense layer(s) in the MLP. 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) in the MLP. The function should return a tf.Tensor.
  • hidden_b_init (callable) – Initializer function for the bias of intermediate dense layer(s) in the MLP. The function should return a tf.Tensor.
  • output_nonlinearity (callable) – Activation function for output dense layer in the MLP. 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) in the MLP. The function should return a tf.Tensor.
  • output_b_init (callable) – Initializer function for the bias of output dense layer(s) in the MLP. The function should return a tf.Tensor.
  • layer_normalization (bool) – Bool for using layer normalization or not.
clone(name)[source]

Return a clone of the Q-function.

It only copies the configuration of the Q-function, not the parameters.

Parameters:name (str) – Name of the newly created q-function.
Returns:Cloned Q function.
Return type:ContinuousCNNQFunction
get_qval(observation, action)[source]

Q Value of the network.

Parameters:
  • observation (np.ndarray) – Observation input of shape \((N, O*)\).
  • action (np.ndarray) – Action input of shape \((N, A*)\).
Returns:

Array of shape \((N, )\) containing Q values

corresponding to each (obs, act) pair.
Return type:

np.ndarray
get_qval_sym(state_input, action_input, name)[source]

Symbolic graph for q-network.

Parameters:
  • state_input (tf.Tensor) – The state input tf.Tensor of shape \((N, O*)\).
  • action_input (tf.Tensor) – The action input tf.Tensor of shape \((N, A*)\).
  • name (str) – Network variable scope.
Returns:

The output Q value tensor of shape \((N, )\).
Return type:

tf.Tensor
inputs

The observation and action input tensors.

The returned tuple contains two tensors. The first is the observation tensor with shape \((N, O*)\), and the second is the action tensor with shape \((N, A*)\).

Type:tuple[tf.Tensor]