Add FNet Backbone

keras_nlp/layers/fnet_encoder.py

@@ -47,6 +47,9 @@ class FNetEncoder(keras.layers.Layer):
             layers.
         bias_initializer: "string" or `keras.initializers` initializer,
             defaults to "zeros". The bias initializer for the dense layers.
+        bias_initializer_output_dense: "string" or `keras.initializers` initializer,
+            defaults to None. The bias initializer for the output dense layer.
+            If None, the bias_initializer will be used.
         name: string, defaults to None. The name of the layer.
         **kwargs: other keyword arguments.
 
@@ -79,6 +82,7 @@ def __init__(
         layer_norm_epsilon=1e-5,
         kernel_initializer="glorot_uniform",
         bias_initializer="zeros",
+        bias_initializer_output_dense=None,
         name=None,
         **kwargs
     ):
@@ -89,6 +93,11 @@ def __init__(
         self.layer_norm_epsilon = layer_norm_epsilon
         self.kernel_initializer = keras.initializers.get(kernel_initializer)
         self.bias_initializer = keras.initializers.get(bias_initializer)
+        if bias_initializer_output_dense is None:
+            bias_initializer_output_dense = bias_initializer
+        self.bias_initializer_output_dense = keras.initializers.get(
+            bias_initializer_output_dense
+        )
 
     def build(self, input_shape):
         # Create layers based on input shape.
@@ -112,7 +121,9 @@ def build(self, input_shape):
         self._output_dense = keras.layers.Dense(
             feature_size,
             kernel_initializer=clone_initializer(self.kernel_initializer),
-            bias_initializer=clone_initializer(self.bias_initializer),
+            bias_initializer=clone_initializer(
+                self.bias_initializer_output_dense
+            ),
         )
         self._output_dropout = keras.layers.Dropout(rate=self.dropout)
 
@@ -170,6 +181,9 @@ def get_config(self):
                 "bias_initializer": keras.initializers.serialize(
                     self.bias_initializer
                 ),
+                "bias_initializer_output_dense": keras.initializers.serialize(
+                    self.bias_initializer_output_dense
+                ),
             }
         )
         return config

keras_nlp/models/__init__.py

@@ -28,6 +28,7 @@
 from keras_nlp.models.distil_bert.distil_bert_tokenizer import (
     DistilBertTokenizer,
 )
+from keras_nlp.models.fnet.fnet_backbone import FNetBackbone
 from keras_nlp.models.roberta.roberta_backbone import RobertaBackbone
 from keras_nlp.models.roberta.roberta_classifier import RobertaClassifier
 from keras_nlp.models.roberta.roberta_preprocessor import RobertaPreprocessor

keras_nlp/models/fnet/__init__.py

@@ -0,0 +1,13 @@
+# Copyright 2022 The KerasNLP Authors
+#
+# 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
+#
+#     https://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.

keras_nlp/models/fnet/fnet_backbone.py

@@ -0,0 +1,214 @@
+# Copyright 2022 The KerasNLP Authors
+#
+# 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
+#
+#     https://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 backbone model."""
+
+import tensorflow as tf
+from tensorflow import keras
+
+from keras_nlp.layers.fnet_encoder import FNetEncoder
+from keras_nlp.layers.position_embedding import PositionEmbedding
+from keras_nlp.models.backbone import Backbone
+
+
+def fnet_kernel_initializer(mode="fnet_default", **kwargs):
+    if mode == "fnet_default":
+        return keras.initializers.RandomNormal(**kwargs)
+    elif mode == "flax_default":
+        return keras.initializers.VarianceScaling(
+            mode="fan_in", distribution="truncated_normal", **kwargs
+        )
+
+
+def fnet_bias_initializer(stddev=0.02):
+    return keras.initializers.RandomNormal(stddev=stddev)
+
+
+@keras.utils.register_keras_serializable(package="keras_nlp")
+class FNetBackbone(Backbone):
+    """FNet encoder network.
+
+    This class implements a bi-directional Fourier Transform-based encoder as
+    described in ["FNet: Mixing Tokens with Fourier Transforms"](https://arxiv.org/abs/2105.03824).
+    It includes the embedding lookups and FNet layers, but not the masked
+    language model or next sentence prediction heads.
+
+    The default constructor gives a fully customizable, randomly initialized FNet
+    encoder with any number of layers and embedding dimensions. To load
+    preset architectures and weights, use the `from_preset` constructor.
+
+    Disclaimer: Pre-trained models are provided on an "as is" basis, without
+    warranties or conditions of any kind.
+
+    Args:
+        vocabulary_size: int. The size of the token vocabulary.
+        num_layers: int. The number of FNet layers.
+        hidden_dim: int. The size of the FNet encoding and pooler layers.
+        intermediate_dim: int. The output dimension of the first Dense layer in
+            a two-layer feedforward network for each FNet layer.
+        dropout: float. Dropout probability for the embeddings and FNet encoder.
+        max_sequence_length: int. The maximum sequence length that this encoder
+            can consume. If None, `max_sequence_length` uses the value from
+            sequence length. This determines the variable shape for positional
+            embeddings.
+        num_segments: int. The number of types that the 'segment_ids' input can
+            take.
+
+    Examples:
+    ```python
+    input_data = {
+        "token_ids": tf.ones(shape=(1, 12), dtype=tf.int64),
+        "segment_ids": tf.constant(
+            [0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0], shape=(1, 12)
+        ),
+    }
+
+    # Randomly initialized FNet encoder with a custom config
+    model = keras_nlp.models.FNetBackbone(
+        vocabulary_size=32000,
+        num_layers=12,
+        hidden_dim=768,
+        intermediate_dim=3072,
+        max_sequence_length=12,
+    )
+    output = model(input_data)
+    ```
+    """
+
+    def __init__(
+        self,
+        vocabulary_size,
+        num_layers,
+        hidden_dim,
+        intermediate_dim,
+        dropout=0.1,
+        max_sequence_length=512,
+        num_segments=4,
+        **kwargs,
+    ):
+
+        # Index of classification token in the vocabulary
+        cls_token_index = 0
+        # Inputs
+        token_id_input = keras.Input(
+            shape=(None,), dtype="int32", name="token_ids"
+        )
+        segment_id_input = keras.Input(
+            shape=(None,), dtype="int32", name="segment_ids"
+        )
+
+        # Embed tokens, positions, and segment ids.
+        token_embedding_layer = keras.layers.Embedding(
+            input_dim=vocabulary_size,
+            output_dim=hidden_dim,
+            embeddings_initializer=fnet_kernel_initializer(stddev=0.02),
+            name="token_embedding",
+        )
+        token_embedding = token_embedding_layer(token_id_input)
+        position_embedding = PositionEmbedding(
+            initializer=fnet_kernel_initializer(stddev=0.02),
+            sequence_length=max_sequence_length,
+            name="position_embedding",
+        )(token_embedding)
+        segment_embedding = keras.layers.Embedding(
+            input_dim=num_segments,
+            output_dim=hidden_dim,
+            embeddings_initializer=fnet_kernel_initializer(stddev=0.02),
+            name="segment_embedding",
+        )(segment_id_input)
+
+        # Sum, normalize and apply dropout to embeddings.
+        x = keras.layers.Add()(
+            (token_embedding, position_embedding, segment_embedding)
+        )
+        x = keras.layers.LayerNormalization(
+            name="embeddings_layer_norm",
+            axis=-1,
+            epsilon=1e-12,
+            dtype=tf.float32,
+        )(x)
+
+        # Project the embedding to `hidden_dim`.
+        x = keras.layers.Dense(
+            hidden_dim,
+            kernel_initializer=fnet_kernel_initializer(
+                "flax_default", scale=1.0
+            ),
+            name="embedding_projection",
+        )(x)
+        x = keras.layers.Dropout(
+            dropout,
+            name="embeddings_dropout",
+        )(x)
+
+        # Apply successive FNet encoder blocks.
+        for i in range(num_layers):
+            x = FNetEncoder(
+                intermediate_dim=intermediate_dim,
+                activation=lambda x: keras.activations.gelu(
+                    x, approximate=True
+                ),
+                layer_norm_epsilon=1e-12,
+                dropout=dropout,
+                kernel_initializer=fnet_kernel_initializer(stddev=0.02),
+                bias_initializer=fnet_bias_initializer(),
+                bias_initializer_output_dense="zeros",
+                name=f"fnet_layer_{i}",
+            )(x)
+
+        # Construct the two FNet outputs. The pooled output is a dense layer on
+        # top of the [CLS] token.
+        sequence_output = x
+        pooled_output = keras.layers.Dense(
+            hidden_dim,
+            kernel_initializer=fnet_kernel_initializer(stddev=0.02),
+            activation="tanh",
+            name="pooled_dense",
+        )(x[:, cls_token_index, :])
+
+        # Instantiate using Functional API Model constructor
+        super().__init__(
+            inputs={
+                "token_ids": token_id_input,
+                "segment_ids": segment_id_input,
+            },
+            outputs={
+                "sequence_output": sequence_output,
+                "pooled_output": pooled_output,
+            },
+            **kwargs,
+        )
+
+        # All references to `self` below this line
+        self.vocabulary_size = vocabulary_size
+        self.num_layers = num_layers
+        self.hidden_dim = hidden_dim
+        self.intermediate_dim = intermediate_dim
+        self.dropout = dropout
+        self.max_sequence_length = max_sequence_length
+        self.num_segments = num_segments
+        self.cls_token_index = cls_token_index
+
+    def get_config(self):
+        return {
+            "vocabulary_size": self.vocabulary_size,
+            "num_layers": self.num_layers,
+            "hidden_dim": self.hidden_dim,
+            "intermediate_dim": self.intermediate_dim,
+            "dropout": self.dropout,
+            "max_sequence_length": self.max_sequence_length,
+            "num_segments": self.num_segments,
+            "name": self.name,
+            "trainable": self.trainable,
+        }