def testDatasetWeighting(self, sizes, weights, target_distribution):
datasets = [
tf.data.Dataset.from_tensor_slices([i] * size)
for i, size in enumerate(sizes)]
if weights is not None:
datasets = (datasets, weights)
dataset = dataset_util.training_pipeline(batch_size=20, shuffle_buffer_size=5000)(datasets)
counts = [0] * len(sizes)
# Check that after 2000 batches we are close to the target distribution.
for x in dataset.take(2000):
for i, _ in enumerate(counts):
counts[i] += int(tf.math.count_nonzero(x == i))
total_count = sum(counts)
for count, freq in zip(counts, target_distribution):
self.assertNear(count / total_count, freq, 0.05)
def make_training_dataset(self,
features_file,
labels_file,
batch_size,
batch_type="examples",
batch_multiplier=1,
batch_size_multiple=1,
shuffle_buffer_size=None,
length_bucket_width=None,
maximum_features_length=None,
maximum_labels_length=None,
single_pass=False,
num_shards=1,
shard_index=0,
num_threads=4,
prefetch_buffer_size=None,
cardinality_multiple=1,
weights=None):
"""See :meth:`opennmt.inputters.ExampleInputter.make_training_dataset`."""
_ = labels_file
dataset = self.make_dataset(features_file, training=True)
if weights is not None:
dataset = (dataset, weights)
transform_fns = []
map_func = lambda x: self.make_features(element=x, training=True)
transform_fns.append(lambda dataset:
dataset.map(map_func,
num_parallel_calls=num_threads or 4))
transform_fns.append(dataset_util.filter_examples_by_length(
maximum_features_length=maximum_features_length,
maximum_labels_length=maximum_labels_length,
features_length_fn=self.get_length))
dataset = dataset_util.training_pipeline(
batch_size,
batch_type=batch_type,
batch_multiplier=batch_multiplier,
batch_size_multiple=batch_size_multiple,
transform_fns=transform_fns,
length_bucket_width=length_bucket_width,
single_pass=single_pass,
num_shards=num_shards,
shard_index=shard_index,
num_threads=num_threads,
shuffle_buffer_size=shuffle_buffer_size,
prefetch_buffer_size=prefetch_buffer_size,
cardinality_multiple=cardinality_multiple)(dataset)
return dataset
def make_training_dataset(self,
features_file,
labels_file,
batch_size,
batch_type="examples",
batch_multiplier=1,
batch_size_multiple=1,
shuffle_buffer_size=None,
length_bucket_width=None,
maximum_features_length=None,
maximum_labels_length=None,
single_pass=False,
num_shards=1,
shard_index=0,
num_threads=4,
prefetch_buffer_size=None):
"""See :meth:`opennmt.inputters.ExampleInputter.make_training_dataset`."""
_ = labels_file
dataset = self.make_dataset(features_file, training=True)
dataset = dataset.apply(
dataset_util.training_pipeline(
batch_size,
batch_type=batch_type,
batch_multiplier=batch_multiplier,
length_bucket_width=length_bucket_width,
single_pass=single_pass,
process_fn=lambda x: self.make_features(element=x,
training=True),
num_threads=num_threads,
shuffle_buffer_size=shuffle_buffer_size,
prefetch_buffer_size=prefetch_buffer_size,
maximum_features_length=maximum_features_length,
maximum_labels_length=maximum_labels_length,
features_length_fn=self.get_length,
batch_size_multiple=batch_size_multiple,
num_shards=num_shards,
shard_index=shard_index))
return dataset
def make_training_dataset(self,
features_file,
labels_file,
batch_size,
batch_type="examples",
batch_multiplier=1,
batch_size_multiple=1,
shuffle_buffer_size=None,
length_bucket_width=None,
maximum_features_length=None,
maximum_labels_length=None,
single_pass=False,
num_shards=1,
shard_index=0,
num_threads=4,
prefetch_buffer_size=None,
cardinality_multiple=1,
weights=None):
"""Builds a dataset to be used for training. It supports the full training
pipeline, including:
* sharding
* shuffling
* filtering
* bucketing
* prefetching
Args:
features_file: The source file or a list of training source files.
labels_file: The target file or a list of training target files.
batch_size: The batch size to use.
batch_type: The training batching stragety to use: can be "examples" or
"tokens".
batch_multiplier: The batch size multiplier to prepare splitting accross
replicated graph parts.
batch_size_multiple: When :obj:`batch_type` is "tokens", ensure that the
resulting batch size is a multiple of this value.
shuffle_buffer_size: The number of elements from which to sample.
length_bucket_width: The width of the length buckets to select batch
candidates from (for efficiency). Set ``None`` to not constrain batch
formation.
maximum_features_length: The maximum length or list of maximum lengths of
the features sequence(s). ``None`` to not constrain the length.
maximum_labels_length: The maximum length of the labels sequence.
``None`` to not constrain the length.
single_pass: If ``True``, makes a single pass over the training data.
num_shards: The number of data shards (usually the number of workers in a
distributed setting).
shard_index: The shard index this data pipeline should read from.
num_threads: The number of elements processed in parallel.
prefetch_buffer_size: The number of batches to prefetch asynchronously. If
``None``, use an automatically tuned value.
cardinality_multiple: Ensure that the dataset cardinality is a multiple of
this value when :obj:`single_pass` is ``True``.
weights: An optional list of weights to create a weighted dataset out of
multiple training files.
Returns:
A ``tf.data.Dataset``.
See Also:
:func:`opennmt.data.training_pipeline`
"""
dataset = self.make_dataset([features_file, labels_file],
training=True)
if weights is not None:
dataset = (dataset, weights)
transform_fns = []
map_func = lambda *arg: self.make_features(element=arg, training=True)
transform_fns.append(lambda dataset: dataset.map(
map_func, num_parallel_calls=num_threads or 4))
transform_fns.append(
dataset_util.filter_examples_by_length(
maximum_features_length=maximum_features_length,
maximum_labels_length=maximum_labels_length,
features_length_fn=self.features_inputter.get_length,
labels_length_fn=self.labels_inputter.get_length))
dataset = dataset_util.training_pipeline(
batch_size,
batch_type=batch_type,
batch_multiplier=batch_multiplier,
batch_size_multiple=batch_size_multiple,
transform_fns=transform_fns,
length_bucket_width=length_bucket_width,
single_pass=single_pass,
num_shards=num_shards,
shard_index=shard_index,
num_threads=num_threads,
shuffle_buffer_size=shuffle_buffer_size,
prefetch_buffer_size=prefetch_buffer_size,
cardinality_multiple=cardinality_multiple)(dataset)
return dataset
def make_training_dataset(
self,
features_file,
labels_file,
batch_size,
batch_type="examples",
batch_multiplier=1,
batch_size_multiple=1,
shuffle_buffer_size=None,
length_bucket_width=None,
maximum_features_length=None,
maximum_labels_length=None,
single_pass=False,
num_shards=1,
shard_index=0,
num_threads=4,
prefetch_buffer_size=None,
cardinality_multiple=1,
weights=None,
batch_autotune_mode=False,
):
"""Builds a dataset to be used for training. It supports the full training
pipeline, including:
* sharding
* shuffling
* filtering
* bucketing
* prefetching
Args:
features_file: The source file or a list of training source files.
labels_file: The target file or a list of training target files.
batch_size: The batch size to use.
batch_type: The training batching strategy to use: can be "examples" or
"tokens".
batch_multiplier: The batch size multiplier to prepare splitting accross
replicated graph parts.
batch_size_multiple: When :obj:`batch_type` is "tokens", ensure that the
resulting batch size is a multiple of this value.
shuffle_buffer_size: The number of elements from which to sample.
length_bucket_width: The width of the length buckets to select batch
candidates from (for efficiency). Set ``None`` to not constrain batch
formation.
maximum_features_length: The maximum length or list of maximum lengths of
the features sequence(s). ``None`` to not constrain the length.
maximum_labels_length: The maximum length of the labels sequence.
``None`` to not constrain the length.
single_pass: If ``True``, makes a single pass over the training data.
num_shards: The number of data shards (usually the number of workers in a
distributed setting).
shard_index: The shard index this data pipeline should read from.
num_threads: The number of elements processed in parallel.
prefetch_buffer_size: The number of batches to prefetch asynchronously. If
``None``, use an automatically tuned value.
cardinality_multiple: Ensure that the dataset cardinality is a multiple of
this value when :obj:`single_pass` is ``True``.
weights: An optional list of weights to create a weighted dataset out of
multiple training files.
batch_autotune_mode: When enabled, all batches are padded to the maximum
sequence length.
Returns:
A ``tf.data.Dataset``.
See Also:
:func:`opennmt.data.training_pipeline`
"""
if labels_file is not None:
data_files = [features_file, labels_file]
maximum_length = [maximum_features_length, maximum_labels_length]
features_length_fn = self.features_inputter.get_length
labels_length_fn = self.labels_inputter.get_length
else:
data_files = features_file
maximum_length = maximum_features_length
features_length_fn = self.get_length
labels_length_fn = None
dataset = self.make_dataset(data_files, training=True)
map_fn = lambda *arg: self.make_features(
element=misc.item_or_tuple(arg), training=True)
filter_fn = lambda *arg: (self.keep_for_training(
misc.item_or_tuple(arg), maximum_length=maximum_length))
transform_fns = [
lambda dataset: dataset.map(map_fn,
num_parallel_calls=num_threads or 4),
lambda dataset: dataset.filter(filter_fn),
]
if batch_autotune_mode:
# In this mode we want to return batches where all sequences are padded
# to the maximum possible length in order to maximize the memory usage.
# Shuffling, sharding, prefetching, etc. are not applied since correctness and
# performance are not important.
if isinstance(dataset, list): # Ignore weighted dataset.
dataset = dataset[0]
# We repeat the dataset now to ensure full batches are always returned.
dataset = dataset.repeat()
for transform_fn in transform_fns:
dataset = dataset.apply(transform_fn)
# length_fn returns the maximum length instead of the actual example length so
# that batches are built as if each example has the maximum length.
if labels_file is not None:
constant_length_fn = [
lambda x: maximum_features_length,
lambda x: maximum_labels_length,
]
else:
constant_length_fn = lambda x: maximum_features_length
# The length dimension is set to the maximum length in the padded shapes.
padded_shapes = self.get_padded_shapes(
dataset.element_spec, maximum_length=maximum_length)
dataset = dataset.apply(
dataset_util.batch_sequence_dataset(
batch_size,
batch_type=batch_type,
batch_multiplier=batch_multiplier,
length_bucket_width=1,
length_fn=constant_length_fn,
padded_shapes=padded_shapes,
))
return dataset
if weights is not None:
dataset = (dataset, weights)
dataset = dataset_util.training_pipeline(
batch_size,
batch_type=batch_type,
batch_multiplier=batch_multiplier,
batch_size_multiple=batch_size_multiple,
transform_fns=transform_fns,
length_bucket_width=length_bucket_width,
features_length_fn=features_length_fn,
labels_length_fn=labels_length_fn,
single_pass=single_pass,
num_shards=num_shards,
shard_index=shard_index,
num_threads=num_threads,
dataset_size=self.get_dataset_size(data_files),
shuffle_buffer_size=shuffle_buffer_size,
prefetch_buffer_size=prefetch_buffer_size,
cardinality_multiple=cardinality_multiple,
)(dataset)
return dataset
