Internal change

PiperOrigin-RevId: 413438617

official/projects/exbert/exbert_pretrain_dataloader.py

+# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Loads dataset for the ExBERT teacher pretraining task."""
+from typing import Mapping, Optional
+
+import tensorflow as tf
+
+from official.core import input_reader
+from official.nlp.data import data_loader
+
+
+class ExbertTeacherPretrainDataLoader(data_loader.DataLoader):
+  """Class to load mixed-vocabulary dataset for Exbert teacher pretrain task."""
+
+  def __init__(self, params):
+    """Inits `ExbertTeacherPretrainDataLoader` class.
+
+    Args:
+      params: A `BertPretrainDataConfig` object.
+    """
+    self._params = params
+    self._seq_length = params.seq_length
+    self._max_predictions_per_seq = params.max_predictions_per_seq
+    self._use_next_sentence_label = params.use_next_sentence_label
+    self._use_position_id = params.use_position_id
+
+  def _name_to_features(self):
+    name_to_features = {
+        'input_mask':
+            tf.io.FixedLenFeature([self._seq_length], tf.int64),
+        'input_mask_teacher':
+            tf.io.FixedLenFeature([self._seq_length], tf.int64),
+        'input_seg_vocab_ids':
+            tf.io.FixedLenFeature([self._seq_length], tf.int64),
+        'masked_lm_positions':
+            tf.io.FixedLenFeature([self._max_predictions_per_seq], tf.int64),
+        'masked_lm_positions_teacher_tvocab':
+            tf.io.FixedLenFeature([self._max_predictions_per_seq], tf.int64),
+        'masked_lm_positions_teacher_svocab':
+            tf.io.FixedLenFeature([self._max_predictions_per_seq], tf.int64),
+        'masked_lm_ids':
+            tf.io.FixedLenFeature([self._max_predictions_per_seq], tf.int64),
+        'masked_lm_ids_teacher_tvocab':
+            tf.io.FixedLenFeature([self._max_predictions_per_seq], tf.int64),
+        'masked_lm_ids_teacher_svocab':
+            tf.io.FixedLenFeature([self._max_predictions_per_seq], tf.int64),
+    }
+    if self._params.use_v2_feature_names:
+      name_to_features.update({
+          'input_word_ids':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+          'input_word_ids_teacher':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+          'input_type_ids':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+          'input_type_ids_teacher':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+      })
+    else:
+      name_to_features.update({
+          'input_ids':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+          'input_ids_teacher':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+          'segment_ids':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+          'segment_ids_teacher':
+              tf.io.FixedLenFeature([self._seq_length], tf.int64),
+      })
+    if self._use_next_sentence_label:
+      name_to_features['next_sentence_labels'] = tf.io.FixedLenFeature([1],
+                                                                       tf.int64)
+    if self._use_position_id:
+      name_to_features['position_ids'] = tf.io.FixedLenFeature(
+          [self._seq_length], tf.int64)
+    return name_to_features
+
+  def _decode(self, record: tf.Tensor):
+    """Decodes a serialized tf.Example."""
+    name_to_features = self._name_to_features()
+    example = tf.io.parse_single_example(record, name_to_features)
+
+    # tf.Example only supports tf.int64, but the TPU only supports tf.int32.
+    # So cast all int64 to int32.
+    for name in list(example.keys()):
+      t = example[name]
+      if t.dtype == tf.int64:
+        t = tf.cast(t, tf.int32)
+      example[name] = t
+
+    return example
+
+  def _parse(self, record: Mapping[str, tf.Tensor]):
+    """Parses raw tensors into a dict of tensors to be consumed by the model."""
+    x = {
+        'input_mask':
+            record['input_mask'],
+        'input_mask_teacher':
+            record['input_mask_teacher'],
+        'input_seg_vocab_ids':
+            record['input_seg_vocab_ids'],
+        'masked_lm_positions':
+            record['masked_lm_positions'],
+        'masked_lm_positions_teacher_tvocab':
+            record['masked_lm_positions_teacher_tvocab'],
+        'masked_lm_positions_teacher_svocab':
+            record['masked_lm_positions_teacher_svocab'],
+        'masked_lm_ids':
+            record['masked_lm_ids'],
+        'masked_lm_ids_teacher_tvocab':
+            record['masked_lm_ids_teacher_tvocab'],
+        'masked_lm_ids_teacher_svocab':
+            record['masked_lm_ids_teacher_svocab'],
+    }
+    if self._params.use_v2_feature_names:
+      x['input_word_ids'] = record['input_word_ids']
+      x['input_word_ids_teacher'] = record['input_word_ids_teacher']
+      x['input_type_ids'] = record['input_type_ids']
+      x['input_type_ids_teacher'] = record['input_type_ids_teacher']
+    else:
+      x['input_word_ids'] = record['input_ids']
+      x['input_word_ids_teacher'] = record['input_ids_teacher']
+      x['input_type_ids'] = record['segment_ids']
+      x['input_type_ids_teacher'] = record['segment_ids_teacher']
+    if self._use_next_sentence_label:
+      x['next_sentence_labels'] = record['next_sentence_labels']
+    if self._use_position_id:
+      x['position_ids'] = record['position_ids']
+
+    return x
+
+  def load(self, input_context: Optional[tf.distribute.InputContext] = None):
+    """Returns a tf.dataset.Dataset."""
+    reader = input_reader.InputReader(
+        params=self._params, decoder_fn=self._decode, parser_fn=self._parse)
+    return reader.read(input_context)

official/projects/exbert/exbert_pretrain_dataloader_test.py

+# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Tests for official.nlp.projects.exbert.exbert_pretrain_dataloader."""
+import itertools
+import os
+
+from absl.testing import parameterized
+import numpy as np
+import tensorflow as tf
+from official.nlp.data import pretrain_dataloader
+from official.projects.exbert import exbert_pretrain_dataloader
+
+
+def create_int_feature(values):
+  f = tf.train.Feature(int64_list=tf.train.Int64List(value=list(values)))
+  return f
+
+
+def _create_fake_exbert_teacher_dataset(output_path,
+                                        seq_length,
+                                        max_predictions_per_seq,
+                                        use_position_id,
+                                        use_next_sentence_label,
+                                        use_v2_feature_names=False):
+  """Creates a fake dataset."""
+  writer = tf.io.TFRecordWriter(output_path)
+
+  for _ in range(100):
+    features = {}
+    input_ids = np.random.randint(100, size=(seq_length))
+    features["input_mask"] = create_int_feature(np.ones_like(input_ids))
+    features["input_mask_teacher"] = create_int_feature(np.ones_like(input_ids))
+    features["input_seg_vocab_ids"] = create_int_feature(
+        np.ones_like(input_ids))
+    if use_v2_feature_names:
+      features["input_word_ids"] = create_int_feature(input_ids)
+      features["input_word_ids_teacher"] = create_int_feature(input_ids)
+      features["input_type_ids"] = create_int_feature(np.ones_like(input_ids))
+      features["input_type_ids_teacher"] = create_int_feature(
+          np.ones_like(input_ids))
+    else:
+      features["input_ids"] = create_int_feature(input_ids)
+      features["input_ids_teacher"] = create_int_feature(input_ids)
+      features["segment_ids"] = create_int_feature(np.ones_like(input_ids))
+      features["segment_ids_teacher"] = create_int_feature(
+          np.ones_like(input_ids))
+
+    features["masked_lm_positions"] = create_int_feature(
+        np.random.randint(100, size=(max_predictions_per_seq)))
+    features["masked_lm_positions_teacher_tvocab"] = create_int_feature(
+        np.random.randint(100, size=(max_predictions_per_seq)))
+    features["masked_lm_positions_teacher_svocab"] = create_int_feature(
+        np.random.randint(100, size=(max_predictions_per_seq)))
+    features["masked_lm_ids"] = create_int_feature(
+        np.random.randint(100, size=(max_predictions_per_seq)))
+    features["masked_lm_ids_teacher_tvocab"] = create_int_feature(
+        np.random.randint(100, size=(max_predictions_per_seq)))
+    features["masked_lm_ids_teacher_svocab"] = create_int_feature(
+        np.random.randint(100, size=(max_predictions_per_seq)))
+
+    if use_next_sentence_label:
+      features["next_sentence_labels"] = create_int_feature([1])
+
+    if use_position_id:
+      features["position_ids"] = create_int_feature(range(0, seq_length))
+
+    tf_example = tf.train.Example(features=tf.train.Features(feature=features))
+    writer.write(tf_example.SerializeToString())
+  writer.close()
+
+
+class ExbertTeacherPretrainDataTest(tf.test.TestCase, parameterized.TestCase):
+
+  @parameterized.parameters(itertools.product(
+      (False, True),
+      (False, True),
+  ))
+  def test_load_data(self, use_next_sentence_label, use_position_id):
+    train_data_path = os.path.join(self.get_temp_dir(), "train.tf_record")
+    seq_length = 128
+    max_predictions_per_seq = 20
+    _create_fake_exbert_teacher_dataset(
+        train_data_path,
+        seq_length,
+        max_predictions_per_seq,
+        use_next_sentence_label=use_next_sentence_label,
+        use_position_id=use_position_id)
+    data_config = pretrain_dataloader.BertPretrainDataConfig(
+        input_path=train_data_path,
+        max_predictions_per_seq=max_predictions_per_seq,
+        seq_length=seq_length,
+        global_batch_size=10,
+        is_training=True,
+        use_next_sentence_label=use_next_sentence_label,
+        use_position_id=use_position_id)
+
+    dataset = exbert_pretrain_dataloader.ExbertTeacherPretrainDataLoader(
+        data_config).load()
+    features = next(iter(dataset))
+    self.assertLen(features,
+                   13 + int(use_next_sentence_label) + int(use_position_id))
+    self.assertIn("input_word_ids", features)
+    self.assertIn("input_word_ids_teacher", features)
+    self.assertIn("input_mask", features)
+    self.assertIn("input_mask_teacher", features)
+    self.assertIn("input_seg_vocab_ids", features)
+    self.assertIn("input_type_ids", features)
+    self.assertIn("input_type_ids_teacher", features)
+    self.assertIn("masked_lm_positions", features)
+    self.assertIn("masked_lm_positions_teacher_tvocab", features)
+    self.assertIn("masked_lm_positions_teacher_svocab", features)
+    self.assertIn("masked_lm_ids", features)
+    self.assertIn("masked_lm_ids_teacher_tvocab", features)
+    self.assertIn("masked_lm_ids_teacher_svocab", features)
+
+    self.assertEqual("next_sentence_labels" in features,
+                     use_next_sentence_label)
+    self.assertEqual("position_ids" in features, use_position_id)
+
+
+if __name__ == "__main__":
+  tf.test.main()