* \brief Get the data as a matrix, where each "row" is the concatenation of rows, and each "column" is the concatenation of columns and batches * \re

Gets parameters in the RNNBuilder.

Creates a random gumbel vector.

* * \brief Set the internal state of a node (for lstms/grus) * \details For RNNs without internal states (SimpleRNN, GRU...), * this has the

* * \brief Get pointer to the current state * * \return Pointer to the current state */

Updates parameters.

For backward compatibility

* \brief Constructor * * \param m ParameterCollection to be trained * \param learning_rate Initial learning rate * \param beta_1 Moving av

Retrieves one internal value in the tensor. # Panics Panics if the tensor has more than one element.

* \brief Get the pointer for a particular batch * \details Automatically broadcasting if the size is zero * * \param bid Batch id requested

Concatenates expressions

* \brief Disable dropout * \details Do this during testing if you want a deterministic network */

Initializes DyNet. # Panics Panics if an invalid value is specified in `params`.

Looks up parameters.

Picks elements from batches

Creates a random bernoulli vector.

Creates a random normal vector.

Gets pointer to the current state.

return Expression of total loss

Adds lookup parameters.

Loads constant parameter.

Applies 2D convolution operation without bias parameters

* \brief Remove one of the dimensions * \param i index of the dimension to be removed */

Disables dropout.

* \brief Get parameters from command line arguments * \details Parses parameters from `argv` and check for required fields depending on the task *

Gets the gradient for the given expression.

Computes batched hinge loss

Computes batched dimensionwise hinge loss

Resets the internally used computation graph with a new one.

Computes batched negative softmax log likelihood

Visualizes the ComputationGraph.

Restarts the trainer.

Computes softmax