Internal change

PiperOrigin-RevId: 476469203

official/nlp/modeling/layers/__init__.py

@@ -30,6 +30,10 @@ from official.nlp.modeling.layers.kernel_attention import KernelMask
 from official.nlp.modeling.layers.masked_lm import MaskedLM
 from official.nlp.modeling.layers.masked_softmax import MaskedSoftmax
 from official.nlp.modeling.layers.mat_mul_with_margin import MatMulWithMargin
+from official.nlp.modeling.layers.mixing import FourierTransformLayer
+from official.nlp.modeling.layers.mixing import HartleyTransformLayer
+from official.nlp.modeling.layers.mixing import LinearTransformLayer
+from official.nlp.modeling.layers.mixing import MixingMechanism
 from official.nlp.modeling.layers.mobile_bert_layers import MobileBertEmbedding
 from official.nlp.modeling.layers.mobile_bert_layers import MobileBertMaskedLM
 from official.nlp.modeling.layers.mobile_bert_layers import MobileBertTransformer

official/nlp/modeling/layers/mixing.py

@@ -26,6 +26,7 @@ Note: These mixing layers currently only support encoder stacks. Decoder stacks
 can be supported in the future by utilizing the `value` inputs.
 """
 
+import enum
 import functools
 from typing import Callable, Tuple, Union
 
@@ -40,6 +41,19 @@ _Initializer = Union[str, tf.keras.initializers.Initializer]
 default_kernel_initializer = tf.keras.initializers.TruncatedNormal(stddev=2e-2)
 
 
+class MixingMechanism(enum.Enum):
+  """Determines the type of mixing layer.
+
+  Possible options:
+    FOURIER: Fourier Transform mixing.
+    LINEAR: Mixing using dense matrix multiplications with learnable weights.
+    HARTLEY: Hartley Transform mixing.
+  """
+  FOURIER = "fourier"
+  HARTLEY = "hartley"
+  LINEAR = "linear"
+
+
 class MixingLayer(tf.keras.layers.Layer):
   """Mixing layer base class.
 

official/nlp/modeling/networks/README.md

@@ -37,3 +37,8 @@ Generalized Autoregressive Pretraining for Language Understanding"
 (https://arxiv.org/abs/1906.08237). It includes embedding lookups,
 relative position encodings, mask computations, segment matrix computations and
 Transformer XL layers using one or two stream relative self-attention.
+
+* [`FNet`](fnet.py) implements the encoder model from ["FNet: Mixing Tokens with
+Fourier Transforms"](https://aclanthology.org/2022.naacl-main.319/). FNet has
+the same structure as a Transformer encoder, except that all or most of the
+self-attention sublayers are replaced with Fourier sublayers.

official/nlp/modeling/networks/__init__.py

@@ -23,6 +23,7 @@ from official.nlp.modeling.networks.bert_encoder import BertEncoder
 from official.nlp.modeling.networks.bert_encoder import BertEncoderV2
 from official.nlp.modeling.networks.classification import Classification
 from official.nlp.modeling.networks.encoder_scaffold import EncoderScaffold
+from official.nlp.modeling.networks.fnet import FNet
 from official.nlp.modeling.networks.funnel_transformer import FunnelTransformerEncoder
 from official.nlp.modeling.networks.mobile_bert_encoder import MobileBERTEncoder
 from official.nlp.modeling.networks.packed_sequence_embedding import PackedSequenceEmbedding

official/nlp/modeling/networks/fnet.py

+# Copyright 2022 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.
+
+"""FNet encoder network.
+
+Based on ["FNet: Mixing Tokens with Fourier Transforms"]
+(https://aclanthology.org/2022.naacl-main.319/).
+"""
+# pylint: disable=g-classes-have-attributes
+
+from typing import Any, Callable, Optional, Sequence, Union
+from absl import logging
+import tensorflow as tf
+
+from official.modeling import tf_utils
+from official.nlp.modeling import layers
+
+_Activation = Union[str, Callable[..., Any]]
+_Initializer = Union[str, tf.keras.initializers.Initializer]
+
+_approx_gelu = lambda x: tf.keras.activations.gelu(x, approximate=True)
+
+
+class FNet(tf.keras.layers.Layer):
+  """FNet encoder network.
+
+  Based on ["FNet: Mixing Tokens with Fourier Transforms"]
+  (https://aclanthology.org/2022.naacl-main.319/). FNet is an efficient
+  Transformer-like encoder network that replaces self-attention sublayers with
+  Fourier sublayers.
+
+  This implementation defaults to the canonical FNet Base model, but the network
+  also supports more general mixing models (e.g. 'Linear', 'HNet') and hybrid
+  models (e.g. 'FNet-Hybrid') models that use both mixing and self-attention
+  layers.
+
+  Args:
+    vocab_size: The size of the token vocabulary.
+    hidden_size: The size of the transformer hidden layers.
+    num_layers: The number of transformer layers.
+    mixing_mechanism: Type of mixing mechanism used in place of self-attention
+      layers. Defaults to FNet ('Fourier') mixing.
+    use_fft: Only used for spectral mixing mechanims. Determines whether to use
+      Fast Fourier Transform (True) or the Discrete Fourier Transform (DFT)
+      matrix (False; default) to compute the Fourier Transform. See
+      layers.FourierTransformLayer or layers.HartleyTransformLayer for advice.
+    attention_layers: Specifies which layers, if any, should be attention layers
+      in the encoder. The remaining [0, num_layers) setminus attention_layers
+      will use the specified `mixing_mechanism`. If using attention layers, a
+      good rule of thumb is to place them in the final few layers.
+    num_attention_heads: The number of attention heads for each transformer. The
+      hidden size must be divisible by the number of attention heads.
+    max_sequence_length: The maximum sequence length that this encoder can
+      consume. This determines the variable shape for positional embeddings.
+    type_vocab_size: The number of types that the 'type_ids' input can take.
+    inner_dim: The output dimension of the first Dense layer in a two-layer
+      feedforward network for each transformer.
+    inner_activation: The activation for the first Dense layer in a two-layer
+      feedforward network for each transformer.
+    output_dropout: Dropout probability for the post-attention and output
+      dropout.
+    attention_dropout: The dropout rate to use for the attention layers within
+      the transformer layers.
+    initializer: The initializer to use for all weights in this encoder.
+    output_range: The sequence output range, [0, output_range), by slicing the
+      target sequence of the last transformer layer. `None` means the entire
+      target sequence will attend to the source sequence, which yields the full
+      output.
+    embedding_width: The width of the word embeddings. If the embedding width is
+      not equal to hidden size, embedding parameters will be factorized into two
+      matrices in the shape of ['vocab_size', 'embedding_width'] and
+      ['embedding_width', 'hidden_size'] ('embedding_width' is usually much
+      smaller than 'hidden_size').
+    embedding_layer: An optional Layer instance which will be called to generate
+      embeddings for the input word IDs.
+    norm_first: Whether to normalize inputs to attention and intermediate dense
+      layers. If set False, output of attention and intermediate dense layers is
+      normalized.
+    with_dense_inputs: Whether to accept dense embeddings as the input.
+  """
+
+  def __init__(
+      self,
+      vocab_size: int,
+      hidden_size: int = 768,
+      num_layers: int = 12,
+      mixing_mechanism: layers.MixingMechanism = layers.MixingMechanism.FOURIER,
+      use_fft: bool = False,
+      attention_layers: Sequence[int] = (),
+      num_attention_heads: int = 12,
+      max_sequence_length: int = 512,
+      type_vocab_size: int = 16,
+      inner_dim: int = 3072,
+      inner_activation: _Activation = _approx_gelu,
+      output_dropout: float = 0.1,
+      attention_dropout: float = 0.1,
+      initializer: _Initializer = tf.keras.initializers.TruncatedNormal(
+          stddev=0.02),
+      output_range: Optional[int] = None,
+      embedding_width: Optional[int] = None,
+      embedding_layer: Optional[tf.keras.layers.Layer] = None,
+      norm_first: bool = False,
+      with_dense_inputs: bool = False,
+      **kwargs):
+    super().__init__(**kwargs)
+
+    activation = tf.keras.activations.get(inner_activation)
+    initializer = tf.keras.initializers.get(initializer)
+
+    if embedding_width is None:
+      embedding_width = hidden_size
+
+    self._config = {
+        'vocab_size': vocab_size,
+        'hidden_size': hidden_size,
+        'num_layers': num_layers,
+        'mixing_mechanism': mixing_mechanism,
+        'use_fft': use_fft,
+        'attention_layers': attention_layers,
+        'num_attention_heads': num_attention_heads,
+        'max_sequence_length': max_sequence_length,
+        'type_vocab_size': type_vocab_size,
+        'inner_dim': inner_dim,
+        'inner_activation': tf.keras.activations.serialize(activation),
+        'output_dropout': output_dropout,
+        'attention_dropout': attention_dropout,
+        'initializer': tf.keras.initializers.serialize(initializer),
+        'output_range': output_range,
+        'embedding_width': embedding_width,
+        'embedding_layer': embedding_layer,
+        'norm_first': norm_first,
+        'with_dense_inputs': with_dense_inputs,
+    }
+
+    if embedding_layer is None:
+      self._embedding_layer = layers.OnDeviceEmbedding(
+          vocab_size=vocab_size,
+          embedding_width=embedding_width,
+          initializer=tf_utils.clone_initializer(initializer),
+          name='word_embeddings')
+    else:
+      self._embedding_layer = embedding_layer
+
+    self._position_embedding_layer = layers.PositionEmbedding(
+        initializer=tf_utils.clone_initializer(initializer),
+        max_length=max_sequence_length,
+        name='position_embedding')
+
+    self._type_embedding_layer = layers.OnDeviceEmbedding(
+        vocab_size=type_vocab_size,
+        embedding_width=embedding_width,
+        initializer=tf_utils.clone_initializer(initializer),
+        use_one_hot=True,
+        name='type_embeddings')
+
+    self._embedding_norm_layer = tf.keras.layers.LayerNormalization(
+        name='embeddings/layer_norm', axis=-1, epsilon=1e-12, dtype=tf.float32)
+
+    self._embedding_dropout = tf.keras.layers.Dropout(
+        rate=output_dropout, name='embedding_dropout')
+
+    # We project the 'embedding' output to 'hidden_size' if it is not already
+    # 'hidden_size'.
+    self._embedding_projection = None
+    if embedding_width != hidden_size:
+      self._embedding_projection = tf.keras.layers.EinsumDense(
+          '...x,xy->...y',
+          output_shape=hidden_size,
+          bias_axes='y',
+          kernel_initializer=tf_utils.clone_initializer(initializer),
+          name='embedding_projection')
+
+    self._transformer_layers = []
+    for layer in range(num_layers):
+      if layer in attention_layers:
+        mixing_layer = layers.MultiHeadAttention(
+            num_heads=num_attention_heads,
+            key_dim=int(hidden_size // num_attention_heads),
+            dropout=attention_dropout,
+            use_bias=True,
+            kernel_initializer=tf_utils.clone_initializer(initializer),
+            name='self_attention',
+        )
+      else:
+        mixing_layer = self._init_mixing_sublayer(layer)
+
+      block = layers.TransformerScaffold(
+          num_attention_heads=num_attention_heads,
+          inner_dim=inner_dim,
+          inner_activation=inner_activation,
+          attention_cls=mixing_layer,
+          feedforward_cls=None,  # Fallback to default FeedForward class
+          output_dropout=output_dropout,
+          attention_dropout=attention_dropout,
+          norm_first=norm_first,
+          output_range=output_range if layer == num_layers - 1 else None,
+          kernel_initializer=tf_utils.clone_initializer(initializer),
+          name='transformer/layer_%d' % layer)
+      self._transformer_layers.append(block)
+
+    self._attention_mask_layer = layers.SelfAttentionMask(
+        name='self_attention_mask')
+
+    self._pooler_layer = tf.keras.layers.Dense(
+        units=hidden_size,
+        activation='tanh',
+        kernel_initializer=tf_utils.clone_initializer(initializer),
+        name='pooler_transform')
+
+    if with_dense_inputs:
+      self.inputs = dict(
+          input_word_ids=tf.keras.Input(shape=(None,), dtype=tf.int32),
+          input_mask=tf.keras.Input(shape=(None,), dtype=tf.int32),
+          input_type_ids=tf.keras.Input(shape=(None,), dtype=tf.int32),
+          dense_inputs=tf.keras.Input(
+              shape=(None, embedding_width), dtype=tf.float32),
+          dense_mask=tf.keras.Input(shape=(None,), dtype=tf.int32),
+          dense_type_ids=tf.keras.Input(shape=(None,), dtype=tf.int32),
+      )
+    else:
+      self.inputs = dict(
+          input_word_ids=tf.keras.Input(shape=(None,), dtype=tf.int32),
+          input_mask=tf.keras.Input(shape=(None,), dtype=tf.int32),
+          input_type_ids=tf.keras.Input(shape=(None,), dtype=tf.int32))
+
+  def call(self, inputs):
+    word_embeddings = None
+    if isinstance(inputs, dict):
+      word_ids = inputs.get('input_word_ids')
+      mask = inputs.get('input_mask')
+      type_ids = inputs.get('input_type_ids')
+      word_embeddings = inputs.get('input_word_embeddings', None)
+
+      dense_inputs = inputs.get('dense_inputs', None)
+      dense_mask = inputs.get('dense_mask', None)
+      dense_type_ids = inputs.get('dense_type_ids', None)
+    else:
+      raise ValueError('Unexpected inputs type (%s) to %s.' %
+                       (type(inputs), self.__class__))
+
+    if word_embeddings is None:
+      word_embeddings = self._embedding_layer(word_ids)
+
+    if dense_inputs is not None:
+      # Concat the dense embeddings at sequence end.
+      word_embeddings = tf.concat([word_embeddings, dense_inputs], axis=1)
+      type_ids = tf.concat([type_ids, dense_type_ids], axis=1)
+      mask = tf.concat([mask, dense_mask], axis=1)
+
+    # Absolute position embeddings.
+    position_embeddings = self._position_embedding_layer(word_embeddings)
+    type_embeddings = self._type_embedding_layer(type_ids)
+
+    embeddings = word_embeddings + position_embeddings + type_embeddings
+    embeddings = self._embedding_norm_layer(embeddings)
+    embeddings = self._embedding_dropout(embeddings)
+
+    if self._embedding_projection is not None:
+      embeddings = self._embedding_projection(embeddings)
+
+    attention_mask = self._attention_mask_layer(embeddings, mask)
+
+    encoder_outputs = []
+    x = embeddings
+    for layer in self._transformer_layers:
+      x = layer([x, attention_mask])
+      encoder_outputs.append(x)
+
+    last_encoder_output = encoder_outputs[-1]
+    first_token_tensor = last_encoder_output[:, 0, :]
+    pooled_output = self._pooler_layer(first_token_tensor)
+
+    output = dict(
+        sequence_output=encoder_outputs[-1],
+        pooled_output=pooled_output,
+        encoder_outputs=encoder_outputs)
+    return output
+
+  def get_embedding_table(self):
+    return self._embedding_layer.embeddings
+
+  def get_embedding_layer(self):
+    return self._embedding_layer
+
+  def get_config(self):
+    return dict(self._config)
+
+  @property
+  def transformer_layers(self):
+    """List of Transformer layers in the encoder."""
+    return self._transformer_layers
+
+  @property
+  def pooler_layer(self):
+    """The pooler dense layer after the transformer layers."""
+    return self._pooler_layer
+
+  @classmethod
+  def from_config(cls, config, custom_objects=None):
+    if 'embedding_layer' in config and config['embedding_layer'] is not None:
+      warn_string = (
+          'You are reloading a model that was saved with a '
+          'potentially-shared embedding layer object. If you contine to '
+          'train this model, the embedding layer will no longer be shared. '
+          'To work around this, load the model outside of the Keras API.')
+      print('WARNING: ' + warn_string)
+      logging.warn(warn_string)
+
+    return cls(**config)
+
+  def _init_mixing_sublayer(self, layer: int):
+    """Initializes config-dependent mixing sublayer."""
+    if self._config['mixing_mechanism'] == layers.MixingMechanism.FOURIER:
+      mixing_sublayer = layers.FourierTransformLayer(
+          use_fft=self._config['use_fft'], name='fourier_transform')
+    elif self._config['mixing_mechanism'] == layers.MixingMechanism.HARTLEY:
+      mixing_sublayer = layers.HartleyTransformLayer(
+          use_fft=self._config['use_fft'], name='hartley_transform')
+    elif self._config['mixing_mechanism'] == layers.MixingMechanism.LINEAR:
+      mixing_sublayer = layers.LinearTransformLayer(
+          kernel_initializer=tf_utils.clone_initializer(
+              self._config['initializer']),
+          name='linear_transform')
+    else:
+      raise ValueError('Unsupported mixing mechanism: %s' %
+                       self._config['mixing_mechanism'])
+
+    return mixing_sublayer

official/nlp/modeling/networks/fnet_test.py

+# Copyright 2022 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 FNet encoder network."""
+
+from typing import Sequence
+
+from absl.testing import parameterized
+import tensorflow as tf
+
+from official.nlp.modeling import layers
+from official.nlp.modeling.networks import fnet
+
+
+class FNetTest(parameterized.TestCase, tf.test.TestCase):
+
+  def tearDown(self):
+    super(FNetTest, self).tearDown()
+    tf.keras.mixed_precision.set_global_policy("float32")
+
+  @parameterized.named_parameters(
+      ("fnet", layers.MixingMechanism.FOURIER, ()),
+      ("fnet_hybrid", layers.MixingMechanism.FOURIER, (1, 2)),
+      ("hnet", layers.MixingMechanism.HARTLEY, ()),
+      ("hnet_hybrid", layers.MixingMechanism.HARTLEY, (1, 2)),
+      ("linear", layers.MixingMechanism.LINEAR, ()),
+      ("linear_hybrid", layers.MixingMechanism.LINEAR, (0,)),
+      ("bert", layers.MixingMechanism.FOURIER, (0, 1, 2)),
+  )
+  def test_network(self, mixing_mechanism: layers.MixingMechanism,
+                   attention_layers: Sequence[int]):
+    num_layers = 3
+    hidden_size = 32
+    sequence_length = 21
+    test_network = fnet.FNet(
+        vocab_size=100,
+        hidden_size=hidden_size,
+        num_attention_heads=2,
+        num_layers=num_layers,
+        mixing_mechanism=mixing_mechanism,
+        attention_layers=attention_layers)
+
+    # Create the inputs (note that the first dimension is implicit).
+    word_ids = tf.keras.Input(shape=(sequence_length,), dtype=tf.int32)
+    mask = tf.keras.Input(shape=(sequence_length,), dtype=tf.int32)
+    type_ids = tf.keras.Input(shape=(sequence_length,), dtype=tf.int32)
+
+    dict_outputs = test_network(
+        dict(input_word_ids=word_ids, input_mask=mask, input_type_ids=type_ids))
+    data = dict_outputs["sequence_output"]
+    pooled = dict_outputs["pooled_output"]
+
+    self.assertIsInstance(test_network.transformer_layers, list)
+    self.assertLen(test_network.transformer_layers, 3)
+    self.assertIsInstance(test_network.pooler_layer, tf.keras.layers.Dense)
+
+    expected_data_shape = [None, sequence_length, hidden_size]
+    expected_pooled_shape = [None, hidden_size]
+    self.assertAllEqual(expected_data_shape, data.shape.as_list())
+    self.assertAllEqual(expected_pooled_shape, pooled.shape.as_list())
+
+    # The default output dtype is float32.
+    self.assertAllEqual(tf.float32, data.dtype)
+    self.assertAllEqual(tf.float32, pooled.dtype)
+
+  def test_embeddings_as_inputs(self):
+    hidden_size = 32
+    sequence_length = 21
+    test_network = fnet.FNet(
+        vocab_size=100,
+        hidden_size=hidden_size,
+        num_attention_heads=2,
+        num_layers=3)
+
+    # Create the inputs (note that the first dimension is implicit).
+    word_ids = tf.keras.Input(shape=(sequence_length), dtype=tf.int32)
+    mask = tf.keras.Input(shape=(sequence_length,), dtype=tf.int32)
+    type_ids = tf.keras.Input(shape=(sequence_length,), dtype=tf.int32)
+
+    test_network.build(
+        dict(input_word_ids=word_ids, input_mask=mask, input_type_ids=type_ids))
+    embeddings = test_network.get_embedding_layer()(word_ids)
+
+    # Calls with the embeddings.
+    dict_outputs = test_network(
+        dict(
+            input_word_embeddings=embeddings,
+            input_mask=mask,
+            input_type_ids=type_ids))
+    all_encoder_outputs = dict_outputs["encoder_outputs"]
+    pooled = dict_outputs["pooled_output"]
+
+    expected_data_shape = [None, sequence_length, hidden_size]
+    expected_pooled_shape = [None, hidden_size]
+    self.assertLen(all_encoder_outputs, 3)
+    for data in all_encoder_outputs:
+      self.assertAllEqual(expected_data_shape, data.shape.as_list())
+    self.assertAllEqual(expected_pooled_shape, pooled.shape.as_list())
+
+    # The default output dtype is float32.
+    self.assertAllEqual(tf.float32, all_encoder_outputs[-1].dtype)
+    self.assertAllEqual(tf.float32, pooled.dtype)
+
+
+if __name__ == "__main__":
+  tf.test.main()