caffe2.python.rnn_cell.RNNCell Class Reference

Inheritance diagram for caffe2.python.rnn_cell.RNNCell:

Public Member Functions
def	__init__ (self, name=None, forward_only=False, initializer=None)
def	initializer (self)
def	initializer (self, value)
def	scope (self, name)
def	apply_over_sequence (self, model, inputs, seq_lengths=None, initial_states=None, outputs_with_grads=None)
def	apply (self, model, input_t, seq_lengths, states, timestep)
def	apply_override (self, model, input_t, seq_lengths, timestep, extra_inputs=None)
def	prepare_input (self, model, input_blob)
def	get_output_state_index (self)
def	get_state_names (self)
def	get_state_names_override (self)
def	get_output_dim (self)

Public Attributes
	name
	recompute_blobs
	forward_only

Detailed Description

Base class for writing recurrent / stateful operations.

One needs to implement 2 methods: apply_override
and get_state_names_override.

As a result base class will provice apply_over_sequence method, which
allows you to apply recurrent operations over a sequence of any length.

As optional you could add input and output preparation steps by overriding
corresponding methods.

Definition at line 48 of file rnn_cell.py.

Member Function Documentation

def caffe2.python.rnn_cell.RNNCell.apply_override	(		self,
			model,
			input_t,
			seq_lengths,
			timestep,
			extra_inputs = None
	)

A single step of a recurrent network to be implemented by each custom
RNNCell.

model: ModelHelper object new operators would be added to

input_t: singlse input with shape (1, batch_size, input_dim)

seq_lengths: blob containing sequence lengths which would be passed to
LSTMUnit operator

states: previous recurrent states

timestep: current recurrent iteration. Could be used together with
seq_lengths in order to determine, if some shorter sequences
in the batch have already ended.

extra_inputs: list of tuples (input, dim). specifies additional input
which is not subject to prepare_input(). (useful when a cell is a
component of a larger recurrent structure, e.g., attention)

Definition at line 192 of file rnn_cell.py.

def caffe2.python.rnn_cell.RNNCell.get_output_dim

(

self

)

Specifies the dimension (number of units) of stepwise output.

Definition at line 251 of file rnn_cell.py.

def caffe2.python.rnn_cell.RNNCell.get_output_state_index

(

self

)

Return index into state list of the "primary" step-wise output.

Definition at line 229 of file rnn_cell.py.

def caffe2.python.rnn_cell.RNNCell.get_state_names

(

self

)

Returns recurrent state names with self.name scoping applied

Definition at line 235 of file rnn_cell.py.

def caffe2.python.rnn_cell.RNNCell.get_state_names_override

(

self

)

Override this function in your custom cell.
It should return the names of the recurrent states.

It's required by apply_over_sequence method in order to allocate
recurrent states for all steps with meaningful names.

Definition at line 241 of file rnn_cell.py.

def caffe2.python.rnn_cell.RNNCell.prepare_input	(		self,
			model,
			input_blob
	)

If some operations in _apply method depend only on the input,
not on recurrent states, they could be computed in advance.

model: ModelHelper object new operators would be added to

input_blob: either the whole input sequence with shape
(sequence_length, batch_size, input_dim) or a single input with shape
(1, batch_size, input_dim).

Definition at line 216 of file rnn_cell.py.

---

The documentation for this class was generated from the following file:

• caffe2/python/rnn_cell.py