tf.contrib.seq2seq.ScheduledOutputTrainingHelper

Class ScheduledOutputTrainingHelper

Inherits From: TrainingHelper

See the guide: Seq2seq Library (contrib) > Dynamic Decoding

A training helper that adds scheduled sampling directly to outputs.

Returns False for sample_ids where no sampling took place; True elsewhere.

Methods

__init__

__init__(
inputs,
sequence_length,
sampling_probability,
time_major=False,
seed=None,
next_inputs_fn=None,
auxiliary_inputs=None,
name=None
)


Initializer.

Args:

• inputs: A (structure) of input tensors.
• sequence_length: An int32 vector tensor.
• sampling_probability: A 0D float32 tensor: the probability of sampling from the outputs instead of reading directly from the inputs.
• time_major: Python bool. Whether the tensors in inputs are time major. If False (default), they are assumed to be batch major.
• seed: The sampling seed.
• next_inputs_fn: (Optional) callable to apply to the RNN outputs to create the next input when sampling. If None (default), the RNN outputs will be used as the next inputs.
• auxiliary_inputs: An optional (structure of) auxiliary input tensors with a shape that matches inputs in all but (potentially) the final dimension. These tensors will be concatenated to the sampled output or the inputs when not sampling for use as the next input.
• name: Name scope for any created operations.

Raises:

• ValueError: if sampling_probability is not a scalar or vector.

initialize

initialize(name=None)


next_inputs

next_inputs(
time,
outputs,
state,
sample_ids,
name=None
)


sample

sample(
time,
outputs,
state,
name=None
)