tokenizers/docs/source/pipeline.rst

The tokenization pipeline
====================================================================================================

When calling :meth:`~tokenizers.Tokenizer.encode` or :meth:`~tokenizers.Tokenizer.encode_batch`, the
input text(s) go through the following pipeline:

- :ref:`normalization`
- :ref:`pre-tokenization`
- :ref:`model`
- :ref:`post-processing`

We'll see in details what happens during each of those steps in detail, as well as when you want to
:ref:`decode <decoding>` some token ids, and how the 🤗 Tokenizers library allows you to customize
each of those steps to your needs. If you're already familiar with those steps and want to learn by
seeing some code, jump to :ref:`our BERT from scratch example <example>`.

For the examples that require a :class:`~tokenizers.Tokenizer`, we will use the tokenizer we trained
in the :doc:`quicktour`, which you can load with:

.. code-block:: python

    from tokenizers import Tokenizer

    tokenizer = Tokenizer.from_file("pretrained/wiki.json")


.. _normalization:

Normalization
----------------------------------------------------------------------------------------------------

Normalization is, in a nutshell, a set of operations you apply to a raw string to make it less
random or "cleaner". Common operations include stripping whitespace, removing accented characters
or lowercasing all text. If you're familiar with `Unicode normalization
<https://unicode.org/reports/tr15>`__, it is also a very common normalization operation applied
in most tokenizers.

Each normalization operation is represented in the 🤗 Tokenizers library by a
:class:`~tokenizers.normalizers.Normalizer`, and you can combine several of those by using a
:class:`~tokenizers.normalizers.Sequence`. Here is a normalizer applying NFD Unicode normalization
and removing accents as an example:

.. code-block:: python

    import tokenizers
    from tokenizers.normalizers import NFD, StripAccents

    normalizer = tokenizers.normalizers.Sequence([NFD(), StripAccents()])

You can apply that normalizer to any string with the
:meth:`~tokenizers.normalizers.Normalizer.normalize_str` method:

.. code-block:: python

    normalizer.normalize_str("Héllò hôw are ü?")
    # "Hello how are u?"

When building a :class:`~tokenizers.Tokenizer`, you can customize its normalizer by just changing
the corresponding attribute:

.. code-block:: python

    tokenizer.normalizer = normalizer

Of course, if you change the way a tokenizer applies normalization, you should probably retrain it
from scratch afterward.


.. _pre-tokenization:

Pre-Tokenization
----------------------------------------------------------------------------------------------------

Pre-tokenization is the act of splitting a text into smaller objects that give an upper bound to
what your tokens will be at the end of training. A good way to think of this is that the
pre-tokenizer will split your text into "words" and then, your final tokens will be parts of those
words.

An easy way to pre-tokenize inputs is to split on spaces and punctuations, which is done by the
:class:`~tokenizers.pre_tokenizers.Whitespace` pre-tokenizer:

.. code-block:: python

    from tokenizers.pre_tokenizers import Whitespace

    pre_tokenizer = Whitespace()
    pre_tokenizer.pre_tokenize_str("Hello! How are you? I'm fine, thank you.")
    # [("Hello", (0, 5)), ("!", (5, 6)), ("How", (7, 10)), ("are", (11, 14)), ("you", (15, 18)),
    #  ("?", (18, 19)), ("I", (20, 21)), ("'", (21, 22)), ('m', (22, 23)), ("fine", (24, 28)),
    #  (",", (28, 29)), ("thank", (30, 35)), ("you", (36, 39)), (".", (39, 40))]

The output is a list of tuples, with each tuple containing one word and its span in the original
sentence (which is used to determine the final :obj:`offsets` of our :class:`~tokenizers.Encoding`).
Note that splitting on punctuation will split contractions like :obj:`"I'm"` in this example.

You can combine together any :class:`~tokenizers.pre_tokenizers.PreTokenizer` together. For
instance, here is a pre-tokenizer that will split on space, punctuation and digits, separating
numbers in their individual digits:

.. code-block:: python

    from tokenizers.pre_tokenizers import Digits

    pre_tokenizer = tokenizers.pre_tokenizers.Sequence([
        Whitespace(),
        Digits(individual_digits=True),
    ])
    pre_tokenizer.pre_tokenize_str("Call 911!")
    # [("Call", (0, 4)), ("9", (5, 6)), ("1", (6, 7)), ("1", (7, 8)), ("!", (8, 9))]

As we saw in the :doc:`quicktour`, you can customize the pre-tokenizer of a
:class:`~tokenizers.Tokenizer` by just changing the corresponding attribute:

.. code-block:: python

    tokenizer.pre_tokenizer = pre_tokenizer

Of course, if you change the way the pre-tokenizer, you should probably retrain your tokenizer from
scratch afterward.


.. _model:

The Model
----------------------------------------------------------------------------------------------------

Once the input texts are normalized and pre-tokenized, we can apply the model on the pre-tokens.
This is the part of the pipeline that needs training on your corpus (or that has been trained if you
are using a pretrained tokenizer).

The role of the model is to split your "words" into tokens, using the rules it has learned. It's
also responsible for mapping those tokens to their corresponding IDs in the vocabulary of the model.

This model is passed along when intializing the :class:`~tokenizers.Tokenizer` so you already know
how to customize this part. Currently, the 🤗 Tokenizers library supports:

- :class:`~tokenizers.models.BPE`
- :class:`~tokenizers.models.Unigram`
- :class:`~tokenizers.models.WordLevel`
- :class:`~tokenizers.models.WordPiece`

For more details about each model and its behavior, you can check `here <components.html#models>`__


.. _post-processing:

Post-Processing
----------------------------------------------------------------------------------------------------

Post-processing is the last step of the tokenization pipeline, to perform any additional
transformation to the :class:`~tokenizers.Encoding` before it's returned, like adding potential
special tokens.

As we saw in the quick tour, we can customize the post processor of a :class:`~tokenizers.Tokenizer`
by setting the corresponding attribute. For instance, here is how we can post-process to make the
inputs suitable for the BERT model:

.. code-block:: python

    from tokenizers.processors import TemplateProcessing

    tokenizer.post_processor = TemplateProcessing
        single="[CLS] $A [SEP]",
        pair="[CLS] $A [SEP] $B:1 [SEP]:1",
        special_tokens=[("[CLS]", 1), ("[SEP]", 2)],
    )

Note that contrarily to the pre-tokenizer or the normalizer, you don't need to retrain a tokenizer
after changing its post-processor.

.. _example:

All together: a BERT tokenizer from scratch
----------------------------------------------------------------------------------------------------

Let's put all those pieces together to build a BERT tokenizer. First, BERT relies on WordPiece, so
we instantiate a new :class:`~tokenizers.Tokenizer` with this model:

.. code-block:: python

    from tokenizers import Tokenizer
    from tokenizers.models import WordPiece

    bert_tokenizer = Tokenizer(WordPiece())

Then we know that BERT preprocesses texts by removing accents and lowercasing. We also use a unicode
normalizer:

.. code-block:: python

    import tokenizers
    from tokenizers.normalizers import Lowercase, NFD, StripAccents

    bert_tokenizer.normalizer = tokenizers.normalizers.Sequence([
        NFD(), Lowercase(), StripAccents()
    ])

The pre-tokenizer is just splitting on whitespace and punctuation:

.. code-block:: python

    from tokenizers.pre_tokenizers import Whitespace

    bert_tokenizer.pre_tokenizer = Whitespace()

And the post-processing uses the template we saw in the previous section:

.. code-block:: python

    from tokenizers.processors import TemplateProcessing

    bert_tokenizer.post_processor = TemplateProcessing(
        single="[CLS] $A [SEP]",
        pair="[CLS] $A [SEP] $B:1 [SEP]:1",
        special_tokens=[
            ("[CLS]", bert_tokenizer.token_to_id("[CLS]")),
            ("[SEP]", bert_tokenizer.token_to_id("[SEP]"))
        ],
    )

We can use this tokenizer and train on it on wikitext like in the :doc:`quicktour`:

.. code-block:: python

    from tokenizers.trainers import WordPieceTrainer

    trainer = WordPieceTrainer(
        vocab_size=30522, special_tokens=["[UNK]", "[CLS]", "[SEP]", "[PAD]", "[MASK]"]
    )
    files = [f"wikitext-103-raw/wiki.{split}.raw" for split in ["test", "train", "valid"]]
    bert_tokenizer.train(trainer, files)

    model_files = bert_tokenizer.model.save("pretrained", "bert-wiki")
    bert_tokenizer.model = WordPiece(*model_files, unk_token="[UNK]")

    bert_tokenizer.save("pretrained/bert-wiki.json")


.. _decoding:

Decoding
----------------------------------------------------------------------------------------------------

On top of encoding the input texts, a :class:`~tokenizers.Tokenizer` also has an API for decoding,
that is converting IDs generated by your model back to a text. This is done by the methods
:meth:`~tokenizers.Tokenizer.decode` (for one predicted text) and
:meth:`~tokenizers.Tokenizer.decode_batch` (for a batch of predictions).

The `decoder` will first convert the IDs back to tokens (using the tokenizer's vocabulary) and
remove all special tokens, then join those tokens with spaces:

.. code-block:: python

    output = tokenizer.encode("Hello, y'all! How are you 😁 ?")
    print(output.ids)
    # [27194, 16, 93, 11, 5068, 5, 7928, 5083, 6190, 0, 35]

    tokenizer.decode([27194, 16, 93, 11, 5068, 5, 7928, 5083, 6190, 0, 35])
    # "Hello , y ' all ! How are you ?"

If you used a model that added special characters to represent subtokens of a given "word" (like
the :obj:`"##"` in WordPiece) you will need to customize the `decoder` to treat them properly. If we
take our previous :obj:`bert_tokenizer` for instance the default decoing will give:

.. code-block:: python

    output = bert_tokenizer.encode("Welcome to the 🤗 Tokenizers library.")
    print(output.tokens)
    # ["[CLS]", "welcome", "to", "the", "[UNK]", "tok", "##eni", "##zer", "##s", "library", ".", "[SEP]"]

    bert_tokenizer.decoder(output.ids)
    # "welcome to the tok ##eni ##zer ##s library ."

But by changing it to a proper decoder, we get:

.. code-block:: python

    bert_tokenizer.decoder = tokenizers.decoders.WordPiece()
    bert_tokenizer.decode(output.ids)
    # "welcome to the tokenizers library."