this wraps an RNN cell to take the context and state and maps it to a fixed diension More...
Inheritance diagram for nabu.neuralnetworks.components.rnn_cell.AttentionProjectionWrapper:
Public Member Functions | |
| def | __init__ (self, cell, output_dim, activation=None) |
| AttentionProjectionWrapper constructor takes the context and the state and maps to output_dim units. More... | |
| def | output_size (self) |
| return cell output size | |
| def | state_size (self) |
| return cell state size | |
| def | zero_state (self, batch_size, dtype) |
| the cell zero state | |
| def | __call__ (self, inputs, state, scope=None) |
| call wrapped cell with constant scope | |
Detailed Description
this wraps an RNN cell to take the context and state and maps it to a fixed diension
Constructor & Destructor Documentation
§ __init__()
| def nabu.neuralnetworks.components.rnn_cell.AttentionProjectionWrapper.__init__ | ( | self, | |
| cell, | |||
| output_dim, | |||
activation = None |
|||
| ) |
AttentionProjectionWrapper constructor takes the context and the state and maps to output_dim units.
- Parameters
-
cell the wrapped cell output_dim the number of output units activation a nonlinearity to be applied after the mapping (default: linear)
The documentation for this class was generated from the following file:
