Input data to RNN
Vector of all RNN trainable parameters concatenated
initial hidden state of the RNN
initial cell state for LSTM networks (only for LSTM)
Vector of valid sequence lengths for each element in batch. (Only used if use_sequence_length kwarg is True)
size of the state for each layer
number of stacked layers
the type of RNN to compute
whether to use bidirectional recurrent layers
Maximum clip value of LSTM states. This option must be used together with lstm_state_clip_min.
Minimum clip value of LSTM states. This option must be used together with lstm_state_clip_max.
Whether to stop NaN from propagating in state by clipping it to min/max. If clipping range is not specified, this option is ignored.
drop rate of the dropout on the outputs of each RNN layer, except the last layer.
size of project size
Whether to have the states as symbol outputs.
If set to true, this layer takes in an extra input parameter
sequence_length to specify variable length sequence