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 |
---|---|
|
int, number of stacked layers |
|
int, number of levels in the data - only used for embedding |
|
int, default = NULL - no embedding. Dimension of the embedding vectors |
|
int, size of the state in each RNN layer |
|
int, number of output variables in the decoding layer |
|
|
|
Either seq-to-one or one-to-one |
|
Type of RNN cell: either gru or lstm |