hub / github.com/tensorflow/lingvo / ZeroState

Method ZeroState

lingvo/core/steps/rnn_steps.py:51–62 · view source on GitHub ↗

Returns the zero_state for the RNN cell. Args: theta: Variables used by the RNNCell. prepared_inputs: unused. batch_size: An int scalar representing the batch size of per-step inputs. Returns: The zero state of the RNNCell.

(self, theta, prepared_inputs, batch_size)

Source from the content-addressed store, hash-verified

source not stored for this graph (policy: none)

Callers 13

testAttentionStepMethod · 0.45

testAttentionStepMultiSourceSameMethod · 0.45

testAttentionStepMultiSourceSameWithGmmAttentionMethod · 0.45

testAttentionStepMultiSourceDifferentMethod · 0.45

testAttentionBlockStepMethod · 0.45

testRnnStepMethod · 0.45

testRnnStepRecurrentMethod · 0.45

testRnnStackStepMethod · 0.45

testRnnStackStepResidualMethod · 0.45

testRnnStackStepNoContextMethod · 0.45

ZeroStateMethod · 0.45

testEmbeddingStepMethod · 0.45

Calls 1

zero_stateMethod · 0.45

Tested by 12

testAttentionStepMethod · 0.36

testAttentionStepMultiSourceSameMethod · 0.36

testAttentionStepMultiSourceSameWithGmmAttentionMethod · 0.36

testAttentionStepMultiSourceDifferentMethod · 0.36

testAttentionBlockStepMethod · 0.36

testRnnStepMethod · 0.36

testRnnStepRecurrentMethod · 0.36

testRnnStackStepMethod · 0.36

testRnnStackStepResidualMethod · 0.36

testRnnStackStepNoContextMethod · 0.36

testEmbeddingStepMethod · 0.36

testStatefulEmbeddingStepMethod · 0.36