rnn.graph

Description

Generate a RNN symbolic model - requires CUDA

Usage

rnn.graph(

  num_rnn_layer,

  input_size = NULL,

  num_embed = NULL,

  num_hidden,

  num_decode,

  dropout = 0,

  ignore_label = -1,

  bidirectional = F,

  loss_output = NULL,

  config,

  cell_type,

  masking = F,

  output_last_state = F,

  rnn.state = NULL,

  rnn.state.cell = NULL,

  prefix = ""

)

Arguments

Argument

Description

num_rnn_layer

int, number of stacked layers

input_size

int, number of levels in the data - only used for embedding

num_embed

int, default = NULL - no embedding. Dimension of the embedding vectors

num_hidden

int, size of the state in each RNN layer

num_decode

int, number of output variables in the decoding layer

dropout

config

Either seq-to-one or one-to-one

cell_type

Type of RNN cell: either gru or lstm