official/projects/lra/mega_encoder.py
# Copyright 2024 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.
"""Mega encoder. Modified From huggingface/transformers."""
# pylint: disable=g-classes-have-attributes
from typing import Any, Callable, Optional, Union
from absl import logging
import tensorflow as tf, tf_keras
import tensorflow_models as tfm
from official.modeling import tf_utils
from official.projects.lra.moving_average_gated_attention import MovingAverageGatedAttention
layers = tfm.nlp.layers
_Initializer = Union[str, tf_keras.initializers.Initializer]
_approx_gelu = lambda x: tf_keras.activations.gelu(x, approximate=True)
@tf_keras.utils.register_keras_serializable(package='Text')
class MegaEncoder(tf_keras.layers.Layer):
"""MegaEncoder.
Args:
vocab_size: The size of the token vocabulary.
embedding_width: The number of embedding dimensions.
intermediate_size: The number of dimension for MLP layers.
num_layers: The number of transformer layers.
max_sequence_length: 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.
type_vocab_size: The number of types that the 'type_ids' input can take.
zdim: hidden dimension for gates used in MEGA Layer.
hdim: hidden dimension used in MEGA Layer.
ndim: number of EMA used in MEGA layer.
activation: The activation for the first Dense layer in a two-layer
feedforward network for each transformer.
bidirectional: Whether to use bidirectional EMA.
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.
hidden_dropout: The dropout rate to use for hidden states in MEGA.
inner_activation: The activation for the first Dense layer in a two-layer
feedforward network for each transformer.
initializer: The initialzer 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_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.
"""
def __init__(
self,
vocab_size: int,
embedding_width: int = 128,
intermediate_size: int = 256,
num_layers: int = 12,
max_sequence_length: int = 512,
type_vocab_size: int = 16,
zdim: int = 64,
hdim: int = 256,
ndim: int = 16,
activation='silu',
bidirectional=False,
dropout: float = 0.0,
attention_dropout: float = 0.0,
hidden_dropout: float = 0.0,
inner_activation: Callable[..., Any] = _approx_gelu,
initializer: _Initializer = tf_keras.initializers.TruncatedNormal(
stddev=0.02
),
output_range: Optional[int] = None,
embedding_layer: Optional[tf_keras.layers.Layer] = None,
norm_first: bool = False,
hidden_size: Optional[int] = None,
**kwargs
):
super().__init__(**kwargs)
# Mega args
initializer = tf_keras.initializers.get(initializer)
if embedding_layer is None:
self._embedding_layer = layers.OnDeviceEmbedding(
vocab_size=vocab_size,
embedding_width=embedding_width,
initializer=initializer,
name='word_embeddings',
)
else:
self._embedding_layer = embedding_layer
self._position_embedding_layer = layers.PositionEmbedding(
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=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=dropout, name='embedding_dropout'
)
self._transformer_layers = []
self._attention_mask_layer = layers.SelfAttentionMask(
name='self_attention_mask'
)
for _ in range(num_layers):
layer = MovingAverageGatedAttention(
embed_dim=embedding_width,
zdim=zdim,
hdim=hdim,
ndim=ndim,
intermediate_size=intermediate_size,
inner_activation=inner_activation,
dropout=dropout,
attention_dropout=attention_dropout,
hidden_dropout=hidden_dropout,
activation=activation,
bidirectional=bidirectional,
prenorm=norm_first,
max_positions=max_sequence_length,
use_bias=True,
return_attention_scores=False,
kernel_initializer=tf_utils.clone_initializer(initializer),
)
self._transformer_layers.append(layer)
self._num_layers = num_layers
self._pooler_layer = tf_keras.layers.Dense(
units=embedding_width,
activation='silu',
kernel_initializer=initializer,
name='pooler_transform',
)
self._config = {
'vocab_size': vocab_size,
'num_layers': num_layers,
'max_sequence_length': max_sequence_length,
'type_vocab_size': type_vocab_size,
'zdim': zdim,
'hdim': hdim,
'ndim': ndim,
'activation': activation,
'bidirectional': bidirectional,
'dropout': dropout,
'attention_dropout': attention_dropout,
'hidden_dropout': hidden_dropout,
'inner_activation': tf_keras.activations.serialize(inner_activation),
'initializer': tf_keras.initializers.serialize(initializer),
'output_range': output_range,
'embedding_width': embedding_width,
'embedding_layer': embedding_layer,
'norm_first': norm_first,
}
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):
if 'input_word_ids' in inputs.keys():
word_ids = inputs.get('input_word_ids')
mask = inputs.get('input_mask')
type_ids = inputs.get('input_type_ids', None)
word_embeddings = inputs.get('input_word_embeddings', None)
elif 'left_word_ids' in inputs.keys():
word_ids = inputs.get('left_word_ids')
mask = inputs.get('left_mask')
elif 'right_word_ids' in inputs.keys():
word_ids = inputs.get('right_word_ids')
mask = inputs.get('right_mask')
dense_inputs = inputs.get('dense_inputs', None)
dense_mask = inputs.get('dense_mask', None)
elif isinstance(inputs, list):
## Dual Encoder Tasks
word_ids, mask = inputs
type_ids = None
dense_inputs, dense_mask = None, None
else:
raise ValueError('Unexpected inputs type to %s.' % self.__class__)
if type_ids is None:
type_ids = tf.zeros_like(mask)
if word_embeddings is None:
word_embeddings = self._embedding_layer(word_ids)
if dense_inputs is not None:
mask = tf.concat([mask, dense_mask], axis=1)
embeddings = self._embedding_norm_layer(word_embeddings)
embeddings = self._embedding_dropout(embeddings)
encoder_outputs = []
x = embeddings
for l in range(self._num_layers):
if x.shape[0] is None:
pass
else:
x = self._transformer_layers[l]([x, mask])
encoder_outputs.append(x)
last_encoder_output = encoder_outputs[-1]
avg_token_tensor = tf.math.reduce_mean(last_encoder_output, axis=1)
pooled_output = self._pooler_layer(avg_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 _get_embeddings(
self,
word_ids: tf.Tensor,
type_ids: tf.Tensor,
word_embeddings: Optional[tf.Tensor],
dense_inputs: Optional[tf.Tensor],
dense_type_ids: Optional[tf.Tensor],
) -> tf.Tensor:
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)
type_embeddings = self._type_embedding_layer(type_ids)
# absolute position embeddings.
position_embeddings = self._position_embedding_layer(word_embeddings)
return word_embeddings + position_embeddings + type_embeddings