View source on GitHub |
A training sampler that adds scheduled sampling directly to outputs.
Inherits From: TrainingSampler
, Sampler
tfa.seq2seq.ScheduledOutputTrainingSampler(
sampling_probability: tfa.types.TensorLike
,
time_major: bool = False,
seed: Optional[int] = None,
next_inputs_fn: Optional[Callable] = None
)
Returns False for sample_ids where no sampling took place; True elsewhere.
Raises | |
---|---|
ValueError
|
if sampling_probability is not a scalar or vector.
|
Methods
initialize
initialize(
inputs, sequence_length=None, mask=None, auxiliary_inputs=None
)
Initialize the TrainSampler.
Args | |
---|---|
inputs
|
A (structure of) input tensors. |
sequence_length
|
An int32 vector tensor. |
mask
|
A boolean 2D tensor. |
Returns | |
---|---|
(finished, next_inputs), a tuple of two items. The first item is a boolean vector to indicate whether the item in the batch has finished. The second item is the first slide of input data based on the timestep dimension (usually the second dim of the input). |
next_inputs
next_inputs(
time, outputs, state, sample_ids
)
Returns (finished, next_inputs, next_state)
.
sample
sample(
time, outputs, state
)
Returns sample_ids
.