Name scope (also acting as variable scope) for this layer.
RNN cell to use.
Initial state to use for the RNN, which is a structure over tensors with shapes
[batchSize, stateShape(i)(0), stateShape(i)(1), ...]
, where i
corresponds to the
index of the corresponding state. Defaults to a zero state.
Boolean value indicating whether the inputs are provided in time-major format (i.e.,
have shape [time, batch, depth]
) or in batch-major format (i.e., have shape
[batch, time, depth]
).
Number of RNN loop iterations allowed to run in parallel.
If true
, GPU-CPU memory swapping support is enabled for the RNN loop.
Optional INT32
tensor with shape [batchSize]
containing the sequence lengths for
each row in the batch.
RNN cell to use.
Initial state to use for the RNN, which is a structure over tensors with shapes
[batchSize, stateShape(i)(0), stateShape(i)(1), ...]
, where i
corresponds to the
index of the corresponding state.
Initial state to use for the RNN, which is a structure over tensors with shapes
[batchSize, stateShape(i)(0), stateShape(i)(1), ...]
, where i
corresponds to the
index of the corresponding state. Defaults to a zero state.
Name scope (also acting as variable scope) for this layer.
Number of RNN loop iterations allowed to run in parallel.
Optional INT32
tensor with shape [batchSize]
containing the sequence lengths for
each row in the batch.
If true
, GPU-CPU memory swapping support is enabled for the RNN loop.
Boolean value indicating whether the inputs are provided in time-major format (i.e.,
have shape [time, batch, depth]
) or in batch-major format (i.e., have shape
[batch, time, depth]
).
Creates a dynamic RNN layer.
$OpDocRNNDynamicRNN