a speller decoder for the LAS architecture More...
Inheritance diagram for nabu.neuralnetworks.models.ed_decoders.speller.Speller:
Public Member Functions | |
| def | create_cell (self, encoded, encoded_seq_length, is_training) |
| create the rnn cell More... | |
 Public Member Functions inherited from nabu.neuralnetworks.models.ed_decoders.rnn_decoder.RNNDecoder | |
| def | create_cell (self, encoded, encoded_seq_length, is_training) |
| create the rnn cell More... | |
| def | zero_state (self, encoded_dim, batch_size) |
| get the decoder zero state More... | |
| def | __getstate__ (self) |
| getstate | |
| Public Member Functions inherited from nabu.neuralnetworks.models.ed_decoders.ed_decoder.EDDecoder | |
| def | __init__ (self, conf, output_dims, constraint, name=None) |
| EDDecoder constructor. More... | |
| def | __call__ (self, encoded, encoded_seq_length, targets, target_seq_length, is_training) |
| def | zero_state (self, encoded_dim, batch_size) |
| get the decoder zero state More... | |
| def | variables (self) |
| get a list of the models's variables | |
Additional Inherited Members | |
| Public Attributes inherited from nabu.neuralnetworks.models.ed_decoders.ed_decoder.EDDecoder | |
| conf | |
| outputs | |
| output_dims | |
| scope | |
Detailed Description
a speller decoder for the LAS architecture
Member Function Documentation
§ create_cell()
| def nabu.neuralnetworks.models.ed_decoders.speller.Speller.create_cell | ( | self, | |
| encoded, | |||
| encoded_seq_length, | |||
| is_training | |||
| ) |
create the rnn cell
- Parameters
-
encoded the encoded sequences as a [batch_size x max_time x dim] tensor that will be queried with attention set to None if the rnn_cell should be created without the attention part (for zero_state) encoded_seq_length the encoded sequence lengths as a [batch_size] vector is_training bool whether or not the network is in training mode
- Returns
- an RNNCell object
The documentation for this class was generated from the following file:
