transformer

May 19, 2024

Copyright 2022 The TensorFlow Authors.

[1]: #@title 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.

1 Neural machine translation with a Transformer and Keras

View on TensorFlow.org
Run in Google Colab
View source on GitHub
Download notebook
This tutorial demonstrates how to create and train a sequence-to-sequence Transformer model to
translate Portuguese into English. The Transformer was originally proposed in “Attention is all
you need” by Vaswani et al. (2017).
Transformers are deep neural networks that replace CNNs and RNNs with self-attention. Self
attention allows Transformers to easily transmit information across the input sequences.
As explained in the Google AI Blog post:
Neural networks for machine translation typically contain an encoder reading the input
sentence and generating a representation of it. A decoder then generates the output
sentence word by word while consulting the representation generated by the encoder.
The Transformer starts by generating initial representations, or embeddings, for each
word… Then, using self-attention, it aggregates information from all of the other words,
generating a new representation per word informed by the entire context, represented
by the filled balls. This step is then repeated multiple times in parallel for all words,
successively generating new representations.

1
Figure 1: Applying the Transformer to machine translation. Source: Google AI Blog.
That’s a lot to digest, the goal of this tutorial is to break it down into easy to understand parts.
In this tutorial you will:
• Prepare the data.
• Implement necessary components:
– Positional embeddings.
– Attention layers.
– The encoder and decoder.
• Build & train the Transformer.
• Generate translations.
• Export the model.
To get the most out of this tutorial, it helps if you know about the basics of text generation and
attention mechanisms.
A Transformer is a sequence-to-sequence encoder-decoder model similar to the model in the NMT
with attention tutorial. A single-layer Transformer takes a little more code to write, but is almost
identical to that encoder-decoder RNN model. The only difference is that the RNN layers are
replaced with self attention layers. This tutorial builds a 4-layer Transformer which is larger and
more powerful, but not fundamentally more complex.
The RNN+Attention model
A 1-layer transformer
After training the model in this notebook, you will be able to input a Portuguese sentence and
return the English translation.
Figure 2: Visualized attention weights that you can generate at the end of this tutorial.

1.1 Why Transformers are significant

• Transformers excel at modeling sequential data, such as natural language.
• Unlike the recurrent neural networks (RNNs), Transformers are parallelizable. This makes
them efficient on hardware like GPUs and TPUs. The main reasons is that Transformers
replaced recurrence with attention, and computations can happen simultaneously. Layer
outputs can be computed in parallel, instead of a series like an RNN.
• Unlike RNNs (like seq2seq, 2014) or convolutional neural networks (CNNs) (for example,
ByteNet), Transformers are able to capture distant or long-range contexts and dependencies in
the data between distant positions in the input or output sequences. Thus, longer connections
can be learned. Attention allows each location to have access to the entire input at each layer,
while in RNNs and CNNs, the information needs to pass through many processing steps to
move a long distance, which makes it harder to learn.
• Transformers make no assumptions about the temporal/spatial relationships across the data.
This is ideal for processing a set of objects (for example, StarCraft units).
Figure 3: The encoder self-attention distribution for the word “it” from the 5th to the 6th layer
of a Transformer trained on English-to-French translation (one of eight attention heads). Source:
Google AI Blog.

2
 1.2 Setup
Begin by installing TensorFlow Datasets for loading the dataset and TensorFlow Text for text
preprocessing:
[2]: # Install the most re version of TensorFlow to use the improved
# masking support for `tf.keras.layers.MultiHeadAttention`.
!apt install --allow-change-held-packages libcudnn8=8.1.0.77-1+cuda11.2
!pip uninstall -y -q tensorflow keras tensorflow-estimator tensorflow-text
!pip install protobuf~=3.20.3
!pip install -q tensorflow_datasets
!pip install -q -U tensorflow-text tensorflow

E: Could not open lock file /var/lib/dpkg/lock-frontend - open (13:

Permission denied)
E: Unable to acquire the dpkg frontend lock (/var/lib/dpkg/lock-
frontend), are you root?
WARNING: Skipping tensorflow-text as it is not installed.
Requirement already satisfied: protobuf~=3.20.3 in
/tmpfs/src/tf_docs_env/lib/python3.9/site-packages (3.20.3)
Import the necessary modules:
[3]: import logging
import time

import numpy as np
import matplotlib.pyplot as plt

import tensorflow_datasets as tfds

import tensorflow as tf

import tensorflow_text

2023-11-16 12:37:14.029081: E
external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register
cuDNN factory: Attempting to register factory for plugin cuDNN when one has
already been registered
2023-11-16 12:37:14.029127: E
external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register
cuFFT factory: Attempting to register factory for plugin cuFFT when one has
already been registered
2023-11-16 12:37:14.030685: E
external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to
register cuBLAS factory: Attempting to register factory for plugin cuBLAS when
one has already been registered

3
 1.3 Data handling
This section downloads the dataset and the subword tokenizer, from this tutorial, then wraps it all
up in a tf.data.Dataset for training.
Toggle section

1.3.1 Download the dataset

Use TensorFlow Datasets to load the Portuguese-English translation datasetD Talks Open Transla-
tion Project. This dataset contains approximately 52,000 training, 1,200 validation and 1,800 test
examples.
[4]: examples, metadata = tfds.load('ted_hrlr_translate/pt_to_en',
with_info=True,
as_supervised=True)

train_examples, val_examples = examples['train'], examples['validation']

The tf.data.Dataset object returned by TensorFlow Datasets yields pairs of text examples:
[5]: for pt_examples, en_examples in train_examples.batch(3).take(1):
print('> Examples in Portuguese:')
for pt in pt_examples.numpy():
print(pt.decode('utf-8'))
print()

print('> Examples in English:')

for en in en_examples.numpy():
print(en.decode('utf-8'))

> Examples in Portuguese:

e quando melhoramos a procura , tiramos a única vantagem da impressão , que é a
serendipidade .
mas e se estes fatores fossem ativos ?
mas eles não tinham a curiosidade de me testar .

> Examples in English:

and when you improve searchability , you actually take away the one advantage of
print , which is serendipity .
but what if it were active ?
but they did n't test for curiosity .

1.3.2 Set up the tokenizer

Now that you have loaded the dataset, you need to tokenize the text, so that each element is
represented as a token or token ID (a numeric representation).
Tokenization is the process of breaking up text, into “tokens”. Depending on the tokenizer, these

4
 tokens can represent sentence-pieces, words, subwords, or characters. To learn more about tok-
enization, visit this guide.
This tutorial uses the tokenizers built in the subword tokenizer tutorial. That tutorial optimizes two
text.BertTokenizer objects (one for English, one for Portuguese) for this dataset and exports
them in a TensorFlow saved_model format.
Note: This is different from the original paper, section 5.1, where they used a single
byte-pair tokenizer for both the source and target with a vocabulary-size of 37000.
Download, extract, and import the saved_model:
[6]: model_name = 'ted_hrlr_translate_pt_en_converter'
tf.keras.utils.get_file(
f'{model_name}.zip',
f'https://storage.googleapis.com/download.tensorflow.org/models/
↪{model_name}.zip',

cache_dir='.', cache_subdir='', extract=True

)

Downloading data from https://storage.googleapis.com/download.tensorflow.org/mod

els/ted_hrlr_translate_pt_en_converter.zip
184801/184801 [==============================] - 0s 0us/step

[6]: './ted_hrlr_translate_pt_en_converter.zip'

[7]: tokenizers = tf.saved_model.load(model_name)

The tf.saved_model contains two text tokenizers, one for English and one for Portuguese. Both
have the same methods:
[8]: [item for item in dir(tokenizers.en) if not item.startswith('_')]

[8]: ['detokenize',
'get_reserved_tokens',
'get_vocab_path',
'get_vocab_size',
'lookup',
'tokenize',
'tokenizer',
'vocab']

The tokenize method converts a batch of strings to a padded-batch of token IDs. This method
splits punctuation, lowercases and unicode-normalizes the input before tokenizing. That standard-
ization is not visible here because the input data is already standardized.
[9]: print('> This is a batch of strings:')
for en in en_examples.numpy():
print(en.decode('utf-8'))

5
 > This is a batch of strings:
and when you improve searchability , you actually take away the one advantage of
print , which is serendipity .
but what if it were active ?
but they did n't test for curiosity .

[10]: encoded = tokenizers.en.tokenize(en_examples)

print('> This is a padded-batch of token IDs:')

for row in encoded.to_list():
print(row)

> This is a padded-batch of token IDs:

[2, 72, 117, 79, 1259, 1491, 2362, 13, 79, 150, 184, 311, 71, 103, 2308, 74,
2679, 13, 148, 80, 55, 4840, 1434, 2423, 540, 15, 3]
[2, 87, 90, 107, 76, 129, 1852, 30, 3]
[2, 87, 83, 149, 50, 9, 56, 664, 85, 2512, 15, 3]
The detokenize method attempts to convert these token IDs back to human-readable text:
[11]: round_trip = tokenizers.en.detokenize(encoded)

print('> This is human-readable text:')

for line in round_trip.numpy():
print(line.decode('utf-8'))

> This is human-readable text:

and when you improve searchability , you actually take away the one advantage of
print , which is serendipity .
but what if it were active ?
but they did n ' t test for curiosity .
The lower level lookup method converts from token-IDs to token text:
[12]: print('> This is the text split into tokens:')
tokens = tokenizers.en.lookup(encoded)
tokens

> This is the text split into tokens:

[12]: <tf.RaggedTensor [[b'[START]', b'and', b'when', b'you', b'improve', b'search',

b'##ability',
b',', b'you', b'actually', b'take', b'away', b'the', b'one', b'advantage',
b'of', b'print', b',', b'which', b'is', b's', b'##ere', b'##nd', b'##ip',
b'##ity', b'.', b'[END]'] ,
[b'[START]', b'but', b'what', b'if', b'it', b'were', b'active', b'?',
b'[END]'] ,
[b'[START]', b'but', b'they', b'did', b'n', b"'", b't', b'test', b'for',
b'curiosity', b'.', b'[END]'] ]>

6
 The output demonstrates the “subword” aspect of the subword tokenization.
For example, the word 'searchability' is decomposed into 'search' and '##ability', and the
word 'serendipity' into 's', '##ere', '##nd', '##ip' and '##ity'.
Note that the tokenized text includes '[START]' and '[END]' tokens.
The distribution of tokens per example in the dataset is as follows:
[13]: lengths = []

for pt_examples, en_examples in train_examples.batch(1024):

pt_tokens = tokenizers.pt.tokenize(pt_examples)
lengths.append(pt_tokens.row_lengths())

en_tokens = tokenizers.en.tokenize(en_examples)
lengths.append(en_tokens.row_lengths())
print('.', end='', flush=True)

[14]: all_lengths = np.concatenate(lengths)

plt.hist(all_lengths, np.linspace(0, 500, 101))

plt.ylim(plt.ylim())
max_length = max(all_lengths)
plt.plot([max_length, max_length], plt.ylim())
plt.title(f'Maximum tokens per example: {max_length}');

7
 1.3.3 Set up a data pipeline with tf.data
The following function takes batches of text as input, and converts them to a format suitable for
training.
1. It tokenizes them into ragged batches.
2. It trims each to be no longer than MAX_TOKENS.
3. It splits the target (English) tokens into inputs and labels. These are shifted by one step so
that at each input location the label is the id of the next token.
4. It converts the RaggedTensors to padded dense Tensors.
5. It returns an (inputs, labels) pair.
[15]: MAX_TOKENS=128
def prepare_batch(pt, en):
pt = tokenizers.pt.tokenize(pt) # Output is ragged.
pt = pt[:, :MAX_TOKENS] # Trim to MAX_TOKENS.
pt = pt.to_tensor() # Convert to 0-padded dense Tensor

en = tokenizers.en.tokenize(en)
en = en[:, :(MAX_TOKENS+1)]
en_inputs = en[:, :-1].to_tensor() # Drop the [END] tokens
en_labels = en[:, 1:].to_tensor() # Drop the [START] tokens

8
 return (pt, en_inputs), en_labels

The function below converts a dataset of text examples into data of batches for training.
1. It tokenizes the text, and filters out the sequences that are too long. (The batch/unbatch is
included because the tokenizer is much more efficient on large batches).
2. The cache method ensures that that work is only executed once.
3. Then shuffle and, dense_to_ragged_batch randomize the order and assemble batches of
examples.
4. Finally prefetch runs the dataset in parallel with the model to ensure that data is available
when needed. See Better performance with the tf.data for details.
[16]: BUFFER_SIZE = 20000
BATCH_SIZE = 64

[17]: def make_batches(ds):

return (
ds
.shuffle(BUFFER_SIZE)
.batch(BATCH_SIZE)
.map(prepare_batch, tf.data.AUTOTUNE)
.prefetch(buffer_size=tf.data.AUTOTUNE))

1.4 Test the Dataset

[18]: # Create training and validation set batches.
train_batches = make_batches(train_examples)
val_batches = make_batches(val_examples)

The resulting tf.data.Dataset objects are setup for training with Keras. Keras Model.fit train-
ing expects (inputs, labels) pairs. The inputs are pairs of tokenized Portuguese and English
sequences, (pt, en). The labels are the same English sequences shifted by 1. This shift is so
that at each location input en sequence, the label in the next token.
Inputs at the bottom, labels at the top.
This is the same as the text generation tutorial, except here you have additional input “context”
(the Portuguese sequence) that the model is “conditioned” on.
This setup is called “teacher forcing” because regardless of the model’s output at each timestep, it
gets the true value as input for the next timestep. This is a simple and efficient way to train a text
generation model. It’s efficient because you don’t need to run the model sequentially, the outputs
at the different sequence locations can be computed in parallel.
You might have expected the input, output, pairs to simply be the Portuguese, English se-
quences. Given the Portuguese sequence, the model would try to generate the English sequence.
It’s possible to train a model that way. You’d need to write out the inference loop and pass the
model’s output back to the input. It’s slower (time steps can’t run in parallel), and a harder task to

9
 learn (the model can’t get the end of a sentence right until it gets the beginning right), but it can
give a more stable model because the model has to learn to correct its own errors during training.
[19]: for (pt, en), en_labels in train_batches.take(1):
break

print(pt.shape)
print(en.shape)
print(en_labels.shape)

(64, 62)
(64, 58)
(64, 58)
The en and en_labels are the same, just shifted by 1:
[20]: print(en[0][:10])
print(en_labels[0][:10])

tf.Tensor([ 2 72 82 76 9 55 154 1664 75 180], shape=(10,),

dtype=int64)
tf.Tensor([ 72 82 76 9 55 154 1664 75 180 6175], shape=(10,),
dtype=int64)

1.5 Define the components

There’s a lot going on inside a Transformer. The important things to remember are:
1. It follows the same general pattern as a standard sequence-to-sequence model with an encoder
and a decoder.
2. If you work through it step by step it will all make sense.
The original Transformer diagram
A representation of a 4-layer Transformer
Each of the components in these two diagrams will be explained as you progress through the
tutorial.

1.5.1 The embedding and positional encoding layer

The inputs to both the encoder and decoder use the same embedding and positional encoding logic.
The embedding and positional encoding layer
Given a sequence of tokens, both the input tokens (Portuguese) and target tokens (English) have
to be converted to vectors using a tf.keras.layers.Embedding layer.
The attention layers used throughout the model see their input as a set of vectors, with no order.
Since the model doesn’t contain any recurrent or convolutional layers. It needs some way to identify
word order, otherwise it would see the input sequence as a bag of words instance, how are you,
how you are, you how are, and so on, are indistinguishable.

10
 A Transformer adds a “Positional Encoding” to the embedding vectors. It uses a set of sines
and cosines at different frequencies (across the sequence). By definition nearby elements will have
similar position encodings.
The original paper uses the following formula for calculating the positional encoding:

𝑃 𝐸(𝑝𝑜𝑠,2𝑖) = sin(𝑝𝑜𝑠/100002𝑖/𝑑𝑚𝑜𝑑𝑒𝑙 )

𝑃 𝐸(𝑝𝑜𝑠,2𝑖+1) = cos(𝑝𝑜𝑠/100002𝑖/𝑑𝑚𝑜𝑑𝑒𝑙 )

Note: The code below implements it, but instead of interleaving the sines and cosines, the vectors
of sines and cosines are simply concatenated. Permuting the channels like this is functionally
equivalent, and just a little easier to implement and show in the plots below.
[21]: def positional_encoding(length, depth):
depth = depth/2

positions = np.arange(length)[:, np.newaxis] # (seq, 1)

depths = np.arange(depth)[np.newaxis, :]/depth # (1, depth)

angle_rates = 1 / (10000**depths) # (1, depth)

angle_rads = positions * angle_rates # (pos, depth)

pos_encoding = np.concatenate(
[np.sin(angle_rads), np.cos(angle_rads)],
axis=-1)

return tf.cast(pos_encoding, dtype=tf.float32)

The position encoding function is a stack of sines and cosines that vibrate at different frequencies
depending on their location along the depth of the embedding vector. They vibrate across the
position axis.
[22]: #@title
pos_encoding = positional_encoding(length=2048, depth=512)

# Check the shape.

print(pos_encoding.shape)

# Plot the dimensions.

plt.pcolormesh(pos_encoding.numpy().T, cmap='RdBu')
plt.ylabel('Depth')
plt.xlabel('Position')
plt.colorbar()
plt.show()

(2048, 512)

11
 By definition these vectors align well with nearby vectors along the position axis. Below the position
encoding vectors are normalized and the vector from position 1000 is compared, by dot-product,
to all the others:
[23]: #@title
pos_encoding/=tf.norm(pos_encoding, axis=1, keepdims=True)
p = pos_encoding[1000]
dots = tf.einsum('pd,d -> p', pos_encoding, p)
plt.subplot(2,1,1)
plt.plot(dots)
plt.ylim([0,1])
plt.plot([950, 950, float('nan'), 1050, 1050],
[0,1,float('nan'),0,1], color='k', label='Zoom')
plt.legend()
plt.subplot(2,1,2)
plt.plot(dots)
plt.xlim([950, 1050])
plt.ylim([0,1])

[23]: (0.0, 1.0)

12
 So use this to create a PositionEmbedding layer that looks-up a token’s embedding vector and
adds the position vector:
[24]: class PositionalEmbedding(tf.keras.layers.Layer):
def __init__(self, vocab_size, d_model):
super().__init__()
self.d_model = d_model
self.embedding = tf.keras.layers.Embedding(vocab_size, d_model,␣
↪mask_zero=True)

self.pos_encoding = positional_encoding(length=2048, depth=d_model)

def compute_mask(self, *args, **kwargs):

return self.embedding.compute_mask(*args, **kwargs)

def call(self, x):

length = tf.shape(x)[1]
x = self.embedding(x)
# This factor sets the relative scale of the embedding and␣
↪positonal_encoding.

x *= tf.math.sqrt(tf.cast(self.d_model, tf.float32))
x = x + self.pos_encoding[tf.newaxis, :length, :]
return x

13
 Note: The original paper, section 3.4 and 5.1, uses a single tokenizer and weight matrix
for both the source and target languages. This tutorial uses two separate tokenizers
and weight matrices.
[25]: embed_pt = PositionalEmbedding(vocab_size=tokenizers.pt.get_vocab_size(),␣
↪d_model=512)

embed_en = PositionalEmbedding(vocab_size=tokenizers.en.get_vocab_size(),␣
↪d_model=512)

pt_emb = embed_pt(pt)
en_emb = embed_en(en)

[26]: en_emb._keras_mask

[26]: <tf.Tensor: shape=(64, 58), dtype=bool, numpy=

array([[ True, True, True, …, False, False, False],
[ True, True, True, …, False, False, False],
[ True, True, True, …, False, False, False],
…,
[ True, True, True, …, False, False, False],
[ True, True, True, …, False, False, False],
[ True, True, True, …, False, False, False]])>

1.5.2 Add and normalize

Add and normalize
These “Add & Norm” blocks are scattered throughout the model. Each one joins a residual con-
nection and runs the result through a LayerNormalization layer.
The easiest way to organize the code is around these residual blocks. The following sections will
define custom layer classes for each.
The residual “Add & Norm” blocks are included so that training is efficient. The residual connection
provides a direct path for the gradient (and ensures that vectors are updated by the attention
layers instead of replaced), while the normalization maintains a reasonable scale for the outputs.
Note: The implementations, below, use the Add layer to ensure that Keras masks are propagated
(the + operator does not).

1.5.3 The base attention layer

Attention layers are used throughout the model. These are all identical except for how the attention
is configured. Each one contains a layers.MultiHeadAttention, a layers.LayerNormalization
and a layers.Add.
The base attention layer
To implement these attention layers, start with a simple base class that just contains the component
layers. Each use-case will be implemented as a subclass. It’s a little more code to write this way,
but it keeps the intention clear.

14
[27]: class BaseAttention(tf.keras.layers.Layer):
def __init__(self, **kwargs):
super().__init__()
self.mha = tf.keras.layers.MultiHeadAttention(**kwargs)
self.layernorm = tf.keras.layers.LayerNormalization()
self.add = tf.keras.layers.Add()

Attention refresher Before you get into the specifics of each usage, here is a quick refresher on
how attention works:
The base attention layer
There are two inputs:
1. The query sequence; the sequence being processed; the sequence doing the attending (bottom).
2. The context sequence; the sequence being attended to (left).
The output has the same shape as the query-sequence.
The common comparison is that this operation is like a dictionary lookup. A fuzzy, differentiable,
vectorized dictionary lookup.
Here’s a regular python dictionary, with 3 keys and 3 values being passed a single query.
d = {'color': 'blue', 'age': 22, 'type': 'pickup'}
result = d['color']
• The querys is what you’re trying to find.
• The keys what sort of information the dictionary has.
• The value is that information.
When you look up a query in a regular dictionary, the dictionary finds the matching key, and
returns its associated value. The query either has a matching key or it doesn’t. You can imagine
a fuzzy dictionary where the keys don’t have to match perfectly. If you looked up d["species"]
in the dictionary above, maybe you’d want it to return "pickup" since that’s the best match for
the query.
An attention layer does a fuzzy lookup like this, but it’s not just looking for the best key. It
combines the values based on how well the query matches each key.
How does that work? In an attention layer the query, key, and value are each vectors. Instead of
doing a hash lookup the attention layer combines the query and key vectors to determine how well
they match, the “attention score”. The layer returns the average across all the values, weighted
by the “attention scores”.
Each location the query-sequence provides a query vector. The context sequence acts as the
dictionary. At each location in the context sequence provides a key and value vector. The input
vectors are not used directly, the layers.MultiHeadAttention layer includes layers.Dense layers
to project the input vectors before using them.

15
 1.5.4 The cross attention layer
At the literal center of the Transformer is the cross-attention layer. This layer connects the encoder
and decoder. This layer is the most straight-forward use of attention in the model, it performs the
same task as the attention block in the NMT with attention tutorial.
The cross attention layer
To implement this you pass the target sequence x as the query and the context sequence as the
key/value when calling the mha layer:
[28]: class CrossAttention(BaseAttention):
def call(self, x, context):
attn_output, attn_scores = self.mha(
query=x,
key=context,
value=context,
return_attention_scores=True)

# Cache the attention scores for plotting later.

self.last_attn_scores = attn_scores

x = self.add([x, attn_output])
x = self.layernorm(x)

return x

The caricature below shows how information flows through this layer. The columns represent the
weighted sum over the context sequence.
For simplicity the residual connections are not shown.
The cross attention layer
The output length is the length of the query sequence, and not the length of the context key/value
sequence.
The diagram is further simplified, below. There’s no need to draw the entire “Attention weights”
matrix. The point is that each query location can see all the key/value pairs in the context, but
no information is exchanged between the queries.
Each query sees the whole context.
Test run it on sample inputs:
[29]: sample_ca = CrossAttention(num_heads=2, key_dim=512)

print(pt_emb.shape)
print(en_emb.shape)
print(sample_ca(en_emb, pt_emb).shape)

(64, 62, 512)

(64, 58, 512)

16
 (64, 58, 512)

1.5.5 The global self attention layer

This layer is responsible for processing the context sequence, and propagating information along
its length:
The global self attention layer
Since the context sequence is fixed while the translation is being generated, information is allowed
to flow in both directions.
Before Transformers and self attention, models commonly used RNNs or CNNs to do this task:
Bidirectional RNNs and CNNs
RNNs and CNNs have their limitations.
• The RNN allows information to flow all the way across the sequence, but it passes through
many processing steps to get there (limiting gradient flow). These RNN steps have to be run
sequentially and so the RNN is less able to take advantage of modern parallel devices.
• In the CNN each location can be processed in parallel, but it only provides a limited receptive
field. The receptive field only grows linearly with the number of CNN layers, You need to
stack a number of Convolution layers to transmit information across the sequence (Wavenet
reduces this problem by using dilated convolutions).
The global self attention layer on the other hand lets every sequence element directly access every
other sequence element, with only a few operations, and all the outputs can be computed in parallel.
To implement this layer you just need to pass the target sequence, x, as both the query, and value
arguments to the mha layer:
[30]: class GlobalSelfAttention(BaseAttention):
def call(self, x):
attn_output = self.mha(
query=x,
value=x,
key=x)
x = self.add([x, attn_output])
x = self.layernorm(x)
return x

[31]: sample_gsa = GlobalSelfAttention(num_heads=2, key_dim=512)

print(pt_emb.shape)
print(sample_gsa(pt_emb).shape)

(64, 62, 512)

(64, 62, 512)
Sticking with the same style as before you could draw it like this:
The global self attention layer

17
 Again, the residual connections are omitted for clarity.
It’s more compact, and just as accurate to draw it like this:
The global self attention layer

1.5.6 The causal self attention layer

This layer does a similar job as the global self attention layer, for the output sequence:
The causal self attention layer
This needs to be handled differently from the encoder’s global self attention layer.
Like the text generation tutorial, and the NMT with attention tutorial, Transformers are an “au-
toregressive” model: They generate the text one token at a time and feed that output back to the
input. To make this efficient, these models ensure that the output for each sequence element only
depends on the previous sequence elements; the models are “causal”.
A single-direction RNN is causal by definition. To make a causal convolution you just need to pad
the input and shift the output so that it aligns correctly (use layers.Conv1D(padding='causal'))
.
Causal RNNs and CNNs
A causal model is efficient in two ways:
1. In training, it lets you compute loss for every location in the output sequence while executing
the model just once.
2. During inference, for each new token generated you only need to calculate its outputs, the
outputs for the previous sequence elements can be reused.
• For an RNN you just need the RNN-state to account for previous computations (pass
return_state=True to the RNN layer’s constructor).
• For a CNN you would need to follow the approach of Fast Wavenet
To build a causal self attention layer, you need to use an appropriate mask when computing the
attention scores and summing the attention values.
This is taken care of automatically if you pass use_causal_mask = True to the
MultiHeadAttention layer when you call it:
[32]: class CausalSelfAttention(BaseAttention):
def call(self, x):
attn_output = self.mha(
query=x,
value=x,
key=x,
use_causal_mask = True)
x = self.add([x, attn_output])
x = self.layernorm(x)
return x

The causal mask ensures that each location only has access to the locations that come before it:

18
 The causal self attention layer
Again, the residual connections are omitted for simplicity.
The more compact representation of this layer would be:
The causal self attention layer
Test out the layer:
[33]: sample_csa = CausalSelfAttention(num_heads=2, key_dim=512)

print(en_emb.shape)
print(sample_csa(en_emb).shape)

(64, 58, 512)

(64, 58, 512)
The output for early sequence elements doesn’t depend on later elements, so it shouldn’t matter if
you trim elements before or after applying the layer:
[34]: out1 = sample_csa(embed_en(en[:, :3]))
out2 = sample_csa(embed_en(en))[:, :3]

tf.reduce_max(abs(out1 - out2)).numpy()

[34]: 4.7683716e-07

Note: When using Keras masks, the output values at invalid locations are not well defined. So the
above may not hold for masked regions.

1.5.7 The feed forward network

The transformer also includes this point-wise feed-forward network in both the encoder and decoder:
The feed forward network
The network consists of two linear layers (tf.keras.layers.Dense) with a ReLU activation in-
between, and a dropout layer. As with the attention layers the code here also includes the residual
connection and normalization:
[35]: class FeedForward(tf.keras.layers.Layer):
def __init__(self, d_model, dff, dropout_rate=0.1):
super().__init__()
self.seq = tf.keras.Sequential([
tf.keras.layers.Dense(dff, activation='relu'),
tf.keras.layers.Dense(d_model),
tf.keras.layers.Dropout(dropout_rate)
])
self.add = tf.keras.layers.Add()
self.layer_norm = tf.keras.layers.LayerNormalization()

19
 def call(self, x):
x = self.add([x, self.seq(x)])
x = self.layer_norm(x)
return x

Test the layer, the output is the same shape as the input:
[36]: sample_ffn = FeedForward(512, 2048)

print(en_emb.shape)
print(sample_ffn(en_emb).shape)

(64, 58, 512)

(64, 58, 512)

1.5.8 The encoder layer

The encoder contains a stack of N encoder layers. Where each EncoderLayer contains a
GlobalSelfAttention and FeedForward layer:
The encoder layer
Here is the definition of the EncoderLayer:
[37]: class EncoderLayer(tf.keras.layers.Layer):
def __init__(self,*, d_model, num_heads, dff, dropout_rate=0.1):
super().__init__()

self.self_attention = GlobalSelfAttention(
num_heads=num_heads,
key_dim=d_model,
dropout=dropout_rate)

self.ffn = FeedForward(d_model, dff)

def call(self, x):

x = self.self_attention(x)
x = self.ffn(x)
return x

And a quick test, the output will have the same shape as the input:
[38]: sample_encoder_layer = EncoderLayer(d_model=512, num_heads=8, dff=2048)

print(pt_emb.shape)
print(sample_encoder_layer(pt_emb).shape)

(64, 62, 512)

(64, 62, 512)

20
 1.5.9 The encoder
Next build the encoder.
The encoder
The encoder consists of:
• A PositionalEmbedding layer at the input.
• A stack of EncoderLayer layers.
[39]: class Encoder(tf.keras.layers.Layer):
def __init__(self, *, num_layers, d_model, num_heads,
dff, vocab_size, dropout_rate=0.1):
super().__init__()

self.d_model = d_model
self.num_layers = num_layers

self.pos_embedding = PositionalEmbedding(
vocab_size=vocab_size, d_model=d_model)

self.enc_layers = [
EncoderLayer(d_model=d_model,
num_heads=num_heads,
dff=dff,
dropout_rate=dropout_rate)
for _ in range(num_layers)]
self.dropout = tf.keras.layers.Dropout(dropout_rate)

def call(self, x):

# `x` is token-IDs shape: (batch, seq_len)
x = self.pos_embedding(x) # Shape `(batch_size, seq_len, d_model)`.

# Add dropout.
x = self.dropout(x)

for i in range(self.num_layers):
x = self.enc_layers[i](x)

return x # Shape `(batch_size, seq_len, d_model)`.

Test the encoder:

[40]: # Instantiate the encoder.
sample_encoder = Encoder(num_layers=4,
d_model=512,
num_heads=8,
dff=2048,
vocab_size=8500)

21
 sample_encoder_output = sample_encoder(pt, training=False)

# Print the shape.

print(pt.shape)
print(sample_encoder_output.shape) # Shape `(batch_size, input_seq_len,␣
↪d_model)`.

(64, 62)
(64, 62, 512)

1.5.10 The decoder layer

The decoder’s stack is slightly more complex, with each DecoderLayer containing a
CausalSelfAttention, a CrossAttention, and a FeedForward layer:
The decoder layer
[41]: class DecoderLayer(tf.keras.layers.Layer):
def __init__(self,
*,
d_model,
num_heads,
dff,
dropout_rate=0.1):
super(DecoderLayer, self).__init__()

self.causal_self_attention = CausalSelfAttention(
num_heads=num_heads,
key_dim=d_model,
dropout=dropout_rate)

self.cross_attention = CrossAttention(
num_heads=num_heads,
key_dim=d_model,
dropout=dropout_rate)

self.ffn = FeedForward(d_model, dff)

def call(self, x, context):

x = self.causal_self_attention(x=x)
x = self.cross_attention(x=x, context=context)

# Cache the last attention scores for plotting later

self.last_attn_scores = self.cross_attention.last_attn_scores

x = self.ffn(x) # Shape `(batch_size, seq_len, d_model)`.

return x

22
 Test the decoder layer:
[42]: sample_decoder_layer = DecoderLayer(d_model=512, num_heads=8, dff=2048)

sample_decoder_layer_output = sample_decoder_layer(
x=en_emb, context=pt_emb)

print(en_emb.shape)
print(pt_emb.shape)
print(sample_decoder_layer_output.shape) # `(batch_size, seq_len, d_model)`

(64, 58, 512)

(64, 62, 512)
(64, 58, 512)

1.5.11 The decoder

Similar to the Encoder, the Decoder consists of a PositionalEmbedding, and a stack of
DecoderLayers:
The embedding and positional encoding layer
Define the decoder by extending tf.keras.layers.Layer:
[43]: class Decoder(tf.keras.layers.Layer):
def __init__(self, *, num_layers, d_model, num_heads, dff, vocab_size,
dropout_rate=0.1):
super(Decoder, self).__init__()

self.d_model = d_model
self.num_layers = num_layers

self.pos_embedding = PositionalEmbedding(vocab_size=vocab_size,
d_model=d_model)
self.dropout = tf.keras.layers.Dropout(dropout_rate)
self.dec_layers = [
DecoderLayer(d_model=d_model, num_heads=num_heads,
dff=dff, dropout_rate=dropout_rate)
for _ in range(num_layers)]

self.last_attn_scores = None

def call(self, x, context):

# `x` is token-IDs shape (batch, target_seq_len)
x = self.pos_embedding(x) # (batch_size, target_seq_len, d_model)

x = self.dropout(x)

for i in range(self.num_layers):

23
 x = self.dec_layers[i](x, context)

self.last_attn_scores = self.dec_layers[-1].last_attn_scores

# The shape of x is (batch_size, target_seq_len, d_model).

return x

Test the decoder:

[44]: # Instantiate the decoder.
sample_decoder = Decoder(num_layers=4,
d_model=512,
num_heads=8,
dff=2048,
vocab_size=8000)

output = sample_decoder(
x=en,
context=pt_emb)

# Print the shapes.

print(en.shape)
print(pt_emb.shape)
print(output.shape)

(64, 58)
(64, 62, 512)
(64, 58, 512)

[45]: sample_decoder.last_attn_scores.shape # (batch, heads, target_seq, input_seq)

[45]: TensorShape([64, 8, 58, 62])

Having created the Transformer encoder and decoder, it’s time to build the Transformer model and
train it.

1.6 The Transformer

You now have Encoder and Decoder. To complete the Transformer model, you need to put them
together and add a final linear (Dense) layer which converts the resulting vector at each location
into output token probabilities.
The output of the decoder is the input to this final linear layer.
The transformer
A Transformer with one layer in both the Encoder and Decoder looks almost exactly like the model
from the RNN+attention tutorial. A multi-layer Transformer has more layers, but is fundamentally
doing the same thing.

24
 A 1-layer transformer
A 4-layer transformer
The RNN+Attention model
Create the Transformer by extending tf.keras.Model:
Note: The original paper, section 3.4, shares the weight matrix between the embedding
layer and the final linear layer. To keep things simple, this tutorial uses two separate
weight matrices.
[46]: class Transformer(tf.keras.Model):
def __init__(self, *, num_layers, d_model, num_heads, dff,
input_vocab_size, target_vocab_size, dropout_rate=0.1):
super().__init__()
self.encoder = Encoder(num_layers=num_layers, d_model=d_model,
num_heads=num_heads, dff=dff,
vocab_size=input_vocab_size,
dropout_rate=dropout_rate)

self.decoder = Decoder(num_layers=num_layers, d_model=d_model,

num_heads=num_heads, dff=dff,
vocab_size=target_vocab_size,
dropout_rate=dropout_rate)

self.final_layer = tf.keras.layers.Dense(target_vocab_size)

def call(self, inputs):

# To use a Keras model with `.fit` you must pass all your inputs in the
# first argument.
context, x = inputs

context = self.encoder(context) # (batch_size, context_len, d_model)

x = self.decoder(x, context) # (batch_size, target_len, d_model)

# Final linear layer output.

logits = self.final_layer(x) # (batch_size, target_len, target_vocab_size)

try:
# Drop the keras mask, so it doesn't scale the losses/metrics.
# b/250038731
del logits._keras_mask
except AttributeError:
pass

# Return the final output and the attention weights.

return logits

25
 1.6.1 Hyperparameters
To keep this example small and relatively fast, the number of layers (num_layers), the dimension-
ality of the embeddings (d_model), and the internal dimensionality of the FeedForward layer (dff)
have been reduced.
The base model described in the original Transformer paper used num_layers=6, d_model=512,
and dff=2048.
The number of self-attention heads remains the same (num_heads=8).

[47]: num_layers = 4
d_model = 128
dff = 512
num_heads = 8
dropout_rate = 0.1

1.6.2 Try it out

Instantiate the Transformer model:
[48]: transformer = Transformer(
num_layers=num_layers,
d_model=d_model,
num_heads=num_heads,
dff=dff,
input_vocab_size=tokenizers.pt.get_vocab_size().numpy(),
target_vocab_size=tokenizers.en.get_vocab_size().numpy(),
dropout_rate=dropout_rate)

Test it:
[49]: output = transformer((pt, en))

print(en.shape)
print(pt.shape)
print(output.shape)

(64, 58)
(64, 62)
(64, 58, 7010)

[50]: attn_scores = transformer.decoder.dec_layers[-1].last_attn_scores

print(attn_scores.shape) # (batch, heads, target_seq, input_seq)

(64, 8, 58, 62)

Print the summary of the model:
[51]: transformer.summary()

26
 Model: "transformer"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
encoder_1 (Encoder) multiple 3632768

decoder_1 (Decoder) multiple 5647104

dense_38 (Dense) multiple 904290

=================================================================
Total params: 10184162 (38.85 MB)
Trainable params: 10184162 (38.85 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

1.7 Training
It’s time to prepare the model and start training it.

1.7.1 Set up the optimizer

Use the Adam optimizer with a custom learning rate scheduler according to the formula in the
original Transformer paper.

−0.5
𝑙𝑟𝑎𝑡𝑒 = 𝑑𝑚𝑜𝑑𝑒𝑙 ∗ min(𝑠𝑡𝑒𝑝_𝑛𝑢𝑚−0.5 , 𝑠𝑡𝑒𝑝_𝑛𝑢𝑚 ⋅ 𝑤𝑎𝑟𝑚𝑢𝑝_𝑠𝑡𝑒𝑝𝑠−1.5 )

[52]: class CustomSchedule(tf.keras.optimizers.schedules.LearningRateSchedule):

def __init__(self, d_model, warmup_steps=4000):
super().__init__()

self.d_model = d_model
self.d_model = tf.cast(self.d_model, tf.float32)

self.warmup_steps = warmup_steps

def __call__(self, step):

step = tf.cast(step, dtype=tf.float32)
arg1 = tf.math.rsqrt(step)
arg2 = step * (self.warmup_steps ** -1.5)

return tf.math.rsqrt(self.d_model) * tf.math.minimum(arg1, arg2)

Instantiate the optimizer (in this example it’s tf.keras.optimizers.Adam):

[53]: learning_rate = CustomSchedule(d_model)

27
 optimizer = tf.keras.optimizers.Adam(learning_rate, beta_1=0.9, beta_2=0.98,
epsilon=1e-9)

Test the custom learning rate scheduler:

[54]: plt.plot(learning_rate(tf.range(40000, dtype=tf.float32)))
plt.ylabel('Learning Rate')
plt.xlabel('Train Step')

[54]: Text(0.5, 0, 'Train Step')

1.7.2 Set up the loss and metrics

Since the target sequences are padded, it is important to apply a padding mask when calculating the
loss. Use the cross-entropy loss function (tf.keras.losses.SparseCategoricalCrossentropy):

[55]: def masked_loss(label, pred):

mask = label != 0
loss_object = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True, reduction='none')
loss = loss_object(label, pred)

28
 mask = tf.cast(mask, dtype=loss.dtype)
loss *= mask

loss = tf.reduce_sum(loss)/tf.reduce_sum(mask)
return loss

def masked_accuracy(label, pred):

pred = tf.argmax(pred, axis=2)
label = tf.cast(label, pred.dtype)
match = label == pred

mask = label != 0

match = match & mask

match = tf.cast(match, dtype=tf.float32)

mask = tf.cast(mask, dtype=tf.float32)
return tf.reduce_sum(match)/tf.reduce_sum(mask)

1.7.3 Train the model

With all the components ready, configure the training procedure using model.compile, and then
run it with model.fit:
Note: This takes about an hour to train in Colab.
[56]: transformer.compile(
loss=masked_loss,
optimizer=optimizer,
metrics=[masked_accuracy])

[57]: transformer.fit(train_batches,
epochs=20,
validation_data=val_batches)

Epoch 1/20
WARNING: All log messages before absl::InitializeLog() is called are written to
STDERR
I0000 00:00:1700138271.584619 39293 device_compiler.h:186] Compiled cluster
using XLA! This line is logged at most once for the lifetime of the process.
810/810 [==============================] - 233s 254ms/step - loss: 6.5998 -
masked_accuracy: 0.1435 - val_loss: 5.0554 - val_masked_accuracy: 0.2485
Epoch 2/20
810/810 [==============================] - 194s 239ms/step - loss: 4.5772 -
masked_accuracy: 0.2972 - val_loss: 4.1541 - val_masked_accuracy: 0.3407
Epoch 3/20

29
810/810 [==============================] - 194s 239ms/step - loss: 3.8242 -
masked_accuracy: 0.3798 - val_loss: 3.5155 - val_masked_accuracy: 0.4196
Epoch 4/20
810/810 [==============================] - 194s 239ms/step - loss: 3.2928 -
masked_accuracy: 0.4375 - val_loss: 3.0246 - val_masked_accuracy: 0.4807
Epoch 5/20
810/810 [==============================] - 193s 238ms/step - loss: 2.9003 -
masked_accuracy: 0.4818 - val_loss: 2.7157 - val_masked_accuracy: 0.5229
Epoch 6/20
810/810 [==============================] - 192s 236ms/step - loss: 2.5720 -
masked_accuracy: 0.5215 - val_loss: 2.5050 - val_masked_accuracy: 0.5463
Epoch 7/20
810/810 [==============================] - 194s 239ms/step - loss: 2.2960 -
masked_accuracy: 0.5575 - val_loss: 2.4090 - val_masked_accuracy: 0.5624
Epoch 8/20
810/810 [==============================] - 193s 237ms/step - loss: 2.1058 -
masked_accuracy: 0.5833 - val_loss: 2.2977 - val_masked_accuracy: 0.5793
Epoch 9/20
810/810 [==============================] - 193s 238ms/step - loss: 1.9573 -
masked_accuracy: 0.6036 - val_loss: 2.1928 - val_masked_accuracy: 0.5937
Epoch 10/20
810/810 [==============================] - 193s 237ms/step - loss: 1.8407 -
masked_accuracy: 0.6204 - val_loss: 2.1310 - val_masked_accuracy: 0.6077
Epoch 11/20
810/810 [==============================] - 194s 239ms/step - loss: 1.7406 -
masked_accuracy: 0.6353 - val_loss: 2.1256 - val_masked_accuracy: 0.6066
Epoch 12/20
810/810 [==============================] - 193s 238ms/step - loss: 1.6556 -
masked_accuracy: 0.6480 - val_loss: 2.1054 - val_masked_accuracy: 0.6122
Epoch 13/20
810/810 [==============================] - 193s 238ms/step - loss: 1.5841 -
masked_accuracy: 0.6585 - val_loss: 2.0783 - val_masked_accuracy: 0.6151
Epoch 14/20
810/810 [==============================] - 190s 234ms/step - loss: 1.5220 -
masked_accuracy: 0.6682 - val_loss: 2.0700 - val_masked_accuracy: 0.6180
Epoch 15/20
810/810 [==============================] - 192s 236ms/step - loss: 1.4666 -
masked_accuracy: 0.6766 - val_loss: 2.0557 - val_masked_accuracy: 0.6194
Epoch 16/20
810/810 [==============================] - 194s 239ms/step - loss: 1.4141 -
masked_accuracy: 0.6855 - val_loss: 2.0391 - val_masked_accuracy: 0.6268
Epoch 17/20
810/810 [==============================] - 193s 238ms/step - loss: 1.3688 -
masked_accuracy: 0.6924 - val_loss: 2.0478 - val_masked_accuracy: 0.6255
Epoch 18/20
810/810 [==============================] - 192s 236ms/step - loss: 1.3287 -
masked_accuracy: 0.6987 - val_loss: 2.0598 - val_masked_accuracy: 0.6251
Epoch 19/20

30
 810/810 [==============================] - 194s 239ms/step - loss: 1.2885 -
masked_accuracy: 0.7053 - val_loss: 2.0598 - val_masked_accuracy: 0.6258
Epoch 20/20
810/810 [==============================] - 193s 238ms/step - loss: 1.2529 -
masked_accuracy: 0.7107 - val_loss: 2.0712 - val_masked_accuracy: 0.6274

[57]: <keras.src.callbacks.History at 0x7f737c5202b0>

1.8 Run inference

You can now test the model by performing a translation. The following steps are used for inference:
• Encode the input sentence using the Portuguese tokenizer (tokenizers.pt). This is the
encoder input.
• The decoder input is initialized to the [START] token.
• Calculate the padding masks and the look ahead masks.
• The decoder then outputs the predictions by looking at the encoder output and its own
output (self-attention).
• Concatenate the predicted token to the decoder input and pass it to the decoder.
• In this approach, the decoder predicts the next token based on the previous tokens it predicted.
Note: The model is optimized for efficient training and makes a next-token prediction for each token
in the output simultaneously. This is redundant during inference, and only the last prediction is
used. This model can be made more efficient for inference if you only calculate the last prediction
when running in inference mode (training=False).
Define the Translator class by subclassing tf.Module:
[58]: class Translator(tf.Module):
def __init__(self, tokenizers, transformer):
self.tokenizers = tokenizers
self.transformer = transformer

def __call__(self, sentence, max_length=MAX_TOKENS):

# The input sentence is Portuguese, hence adding the `[START]` and `[END]`␣
↪tokens.

assert isinstance(sentence, tf.Tensor)

if len(sentence.shape) == 0:
sentence = sentence[tf.newaxis]

sentence = self.tokenizers.pt.tokenize(sentence).to_tensor()

encoder_input = sentence

# As the output language is English, initialize the output with the

# English `[START]` token.
start_end = self.tokenizers.en.tokenize([''])[0]
start = start_end[0][tf.newaxis]
end = start_end[1][tf.newaxis]

31
 # `tf.TensorArray` is required here (instead of a Python list), so that the
# dynamic-loop can be traced by `tf.function`.
output_array = tf.TensorArray(dtype=tf.int64, size=0, dynamic_size=True)
output_array = output_array.write(0, start)

for i in tf.range(max_length):
output = tf.transpose(output_array.stack())
predictions = self.transformer([encoder_input, output], training=False)

# Select the last token from the `seq_len` dimension.

predictions = predictions[:, -1:, :] # Shape `(batch_size, 1,␣
↪vocab_size)`.

predicted_id = tf.argmax(predictions, axis=-1)

# Concatenate the `predicted_id` to the output which is given to the

# decoder as its input.
output_array = output_array.write(i+1, predicted_id[0])

if predicted_id == end:
break

output = tf.transpose(output_array.stack())
# The output shape is `(1, tokens)`.
text = tokenizers.en.detokenize(output)[0] # Shape: `()`.

tokens = tokenizers.en.lookup(output)[0]

# `tf.function` prevents us from using the attention_weights that were

# calculated on the last iteration of the loop.
# So, recalculate them outside the loop.
self.transformer([encoder_input, output[:,:-1]], training=False)
attention_weights = self.transformer.decoder.last_attn_scores

return text, tokens, attention_weights

Note: This function uses an unrolled loop, not a dynamic loop. It generates MAX_TOKENS on every
call. Refer to the NMT with attention tutorial for an example implementation with a dynamic
loop, which can be much more efficient.
Create an instance of this Translator class, and try it out a few times:
[59]: translator = Translator(tokenizers, transformer)

[60]: def print_translation(sentence, tokens, ground_truth):

print(f'{"Input:":15s}: {sentence}')
print(f'{"Prediction":15s}: {tokens.numpy().decode("utf-8")}')

32
 print(f'{"Ground truth":15s}: {ground_truth}')

Example 1:
[61]: sentence = 'este é um problema que temos que resolver.'
ground_truth = 'this is a problem we have to solve .'

translated_text, translated_tokens, attention_weights = translator(

tf.constant(sentence))
print_translation(sentence, translated_text, ground_truth)

Input: : este é um problema que temos que resolver.

Prediction : so this is a problem we have to solve .
Ground truth : this is a problem we have to solve .
Example 2:
[62]: sentence = 'os meus vizinhos ouviram sobre esta ideia.'
ground_truth = 'and my neighboring homes heard about this idea .'

translated_text, translated_tokens, attention_weights = translator(

tf.constant(sentence))
print_translation(sentence, translated_text, ground_truth)

Input: : os meus vizinhos ouviram sobre esta ideia.

Prediction : my neighbors heard about this idea .
Ground truth : and my neighboring homes heard about this idea .
Example 3:
[63]: sentence = 'vou então muito rapidamente partilhar convosco algumas histórias de␣
↪algumas coisas mágicas que aconteceram.'

ground_truth = "so i'll just share with you some stories very quickly of some␣
↪magical things that have happened."

translated_text, translated_tokens, attention_weights = translator(

tf.constant(sentence))
print_translation(sentence, translated_text, ground_truth)

Input: : vou então muito rapidamente partilhar convosco algumas

histórias de algumas coisas mágicas que aconteceram.
Prediction : so i ' m going to share a lot of these things with you some
magic stories that happened .
Ground truth : so i'll just share with you some stories very quickly of some
magical things that have happened.

1.9 Create attention plots

The Translator class you created in the previous section returns a dictionary of attention heatmaps
you can use to visualize the internal working of the model.

33
 For example:
[64]: sentence = 'este é o primeiro livro que eu fiz.'
ground_truth = "this is the first book i've ever done."

translated_text, translated_tokens, attention_weights = translator(

tf.constant(sentence))
print_translation(sentence, translated_text, ground_truth)

Input: : este é o primeiro livro que eu fiz.

Prediction : this is the first book i did .
Ground truth : this is the first book i've ever done.
Create a function that plots the attention when a token is generated:
[65]: def plot_attention_head(in_tokens, translated_tokens, attention):
# The model didn't generate `<START>` in the output. Skip it.
translated_tokens = translated_tokens[1:]

ax = plt.gca()
ax.matshow(attention)
ax.set_xticks(range(len(in_tokens)))
ax.set_yticks(range(len(translated_tokens)))

labels = [label.decode('utf-8') for label in in_tokens.numpy()]

ax.set_xticklabels(
labels, rotation=90)

labels = [label.decode('utf-8') for label in translated_tokens.numpy()]

ax.set_yticklabels(labels)

[66]: head = 0
# Shape: `(batch=1, num_heads, seq_len_q, seq_len_k)`.
attention_heads = tf.squeeze(attention_weights, 0)
attention = attention_heads[head]
attention.shape

[66]: TensorShape([9, 11])

These are the input (Portuguese) tokens:

[67]: in_tokens = tf.convert_to_tensor([sentence])

in_tokens = tokenizers.pt.tokenize(in_tokens).to_tensor()
in_tokens = tokenizers.pt.lookup(in_tokens)[0]
in_tokens

[67]: <tf.Tensor: shape=(11,), dtype=string, numpy=

array([b'[START]', b'este', b'e', b'o', b'primeiro', b'livro', b'que',
b'eu', b'fiz', b'.', b'[END]'], dtype=object)>

34
 And these are the output (English translation) tokens:

[68]: translated_tokens

[68]: <tf.Tensor: shape=(10,), dtype=string, numpy=

array([b'[START]', b'this', b'is', b'the', b'first', b'book', b'i',
b'did', b'.', b'[END]'], dtype=object)>

[69]: plot_attention_head(in_tokens, translated_tokens, attention)

[70]: def plot_attention_weights(sentence, translated_tokens, attention_heads):

in_tokens = tf.convert_to_tensor([sentence])
in_tokens = tokenizers.pt.tokenize(in_tokens).to_tensor()
in_tokens = tokenizers.pt.lookup(in_tokens)[0]

fig = plt.figure(figsize=(16, 8))

for h, head in enumerate(attention_heads):

ax = fig.add_subplot(2, 4, h+1)

35
 plot_attention_head(in_tokens, translated_tokens, head)

ax.set_xlabel(f'Head {h+1}')

plt.tight_layout()
plt.show()

[71]: plot_attention_weights(sentence,
translated_tokens,
attention_weights[0])

The model can handle unfamiliar words. Neither 'triceratops' nor 'encyclopédia' are in the
input dataset, and the model attempts to transliterate them even without a shared vocabulary. For
example:
[72]: sentence = 'Eu li sobre triceratops na enciclopédia.'
ground_truth = 'I read about triceratops in the encyclopedia.'

translated_text, translated_tokens, attention_weights = translator(

tf.constant(sentence))
print_translation(sentence, translated_text, ground_truth)

plot_attention_weights(sentence, translated_tokens, attention_weights[0])

Input: : Eu li sobre triceratops na enciclopédia.

Prediction : i read about chonese chops into egypt .
Ground truth : I read about triceratops in the encyclopedia.

36
 1.10 Export the model
You have tested the model and the inference is working. Next, you can export it as a
tf.saved_model. To learn about saving and loading a model in the SavedModel format, use
this guide.
Create a class called ExportTranslator by subclassing the tf.Module subclass with a tf.function
on the __call__ method:
[73]: class ExportTranslator(tf.Module):
def __init__(self, translator):
self.translator = translator

@tf.function(input_signature=[tf.TensorSpec(shape=[], dtype=tf.string)])
def __call__(self, sentence):
(result,
tokens,
attention_weights) = self.translator(sentence, max_length=MAX_TOKENS)

return result

In the above tf.function only the output sentence is returned. Thanks to the non-strict execution
in tf.function any unnecessary values are never computed.
Wrap translator in the newly created ExportTranslator:
[74]: translator = ExportTranslator(translator)

Since the model is decoding the predictions using tf.argmax the predictions are deterministic. The
original model and one reloaded from its SavedModel should give identical predictions:

37
[75]: translator('este é o primeiro livro que eu fiz.').numpy()

[75]: b'this is the first book i did .'

[76]: tf.saved_model.save(translator, export_dir='translator')

WARNING:tensorflow:Model's `__init__()` arguments contain non-serializable

objects. Please implement a `get_config()` method in the subclassed Model for
proper saving and loading. Defaulting to empty config.
WARNING:tensorflow:Model's `__init__()` arguments contain non-serializable
objects. Please implement a `get_config()` method in the subclassed Model for
proper saving and loading. Defaulting to empty config.
WARNING:tensorflow:Model's `__init__()` arguments contain non-serializable
objects. Please implement a `get_config()` method in the subclassed Model for
proper saving and loading. Defaulting to empty config.
WARNING:tensorflow:Model's `__init__()` arguments contain non-serializable
objects. Please implement a `get_config()` method in the subclassed Model for
proper saving and loading. Defaulting to empty config.
INFO:tensorflow:Assets written to: translator/assets
INFO:tensorflow:Assets written to: translator/assets

[77]: reloaded = tf.saved_model.load('translator')

[78]: reloaded('este é o primeiro livro que eu fiz.').numpy()

[78]: b'this is the first book i did .'

1.11 Conclusion
In this tutorial you learned about:
• The Transformers and their significance in machine learning
• Attention, self-attention and multi-head attention
• Positional encoding with embeddings
• The encoder-decoder architecture of the original Transformer
• Masking in self-attention
• How to put it all together to translate text
The downsides of this architecture are:
• For a time-series, the output for a time-step is calculated from the entire history instead of
only the inputs and current hidden-state. This may be less efficient.
• If the input has a temporal/spatial relationship, like text or images, some positional encoding
must be added or the model will effectively see a bag of words.
If you want to practice, there are many things you could try with it. For example:

38
 • Use a different dataset to train the Transformer.
• Create the “Base Transformer” or “Transformer XL” configurations from the original paper
by changing the hyperparameters.
• Use the layers defined here to create an implementation of BERT
• Use Beam search to get better predictions.
There are a wide variety of Transformer-based models, many of which improve upon the 2017 version
of the original Transformer with encoder-decoder, encoder-only and decoder-only architectures.
Some of these models are covered in the following research publications:
• “Efficient Transformers: a survey” (Tay et al., 2022)
• “Formal algorithms for Transformers” (Phuong and Hutter, 2022).
• T5 (“Exploring the limits of transfer learning with a unified text-to-text Transformer”) (Raffel
et al., 2019)
You can learn more about other models in the following Google blog posts:
• PaLM.
• LaMDA
• MUM
• Reformer
• BERT
If you’re interested in studying how attention-based models have been applied in tasks outside of
natural language processing, check out the following resources:
• Vision Transformer (ViT): Transformers for image recognition at scale
• Multi-task multitrack music transcription (MT3) with a Transformer
• Code generation with AlphaCode
• Reinforcement learning with multi-game decision Transformers
• Protein structure prediction with AlphaFold
• OptFormer: Towards universal hyperparameter optimization with Transformers
[ ]:

39