docs/source-doc-builder/pipeline.mdx
When calling Tokenizer.encode or
Tokenizer.encode_batch, the input
text(s) go through the following pipeline:
normalizationpre-tokenizationmodelpost-processingWe'll see in details what happens during each of those steps in detail,
as well as when you want to 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
our BERT from scratch example <example>.
For the examples that require a Tokenizer we will use the tokenizer we trained in the
quicktour, which you can load with:
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, it is also a very common normalization operation applied in most tokenizers.
Each normalization operation is represented in the 🤗 Tokenizers library
by a Normalizer, and you can combine
several of those by using a normalizers.Sequence. Here is a normalizer applying NFD Unicode normalization
and removing accents as an example:
You can manually test that normalizer by applying it to any string:
<tokenizerslangcontent> <python> <literalinclude> {"path": "../../bindings/python/tests/documentation/test_pipeline.py", "language": "python", "start-after": "START test_normalizer", "end-before": "END test_normalizer", "dedent": 8} </literalinclude> </python> <rust> <literalinclude> {"path": "../../tokenizers/tests/documentation.rs", "language": "rust", "start-after": "START pipeline_test_normalizer", "end-before": "END pipeline_test_normalizer", "dedent": 4} </literalinclude> </rust> <node> <literalinclude> {"path": "../../bindings/node/examples/documentation/pipeline.test.ts", "language": "js", "start-after": "START test_normalizer", "end-before": "END test_normalizer", "dedent": 8} </literalinclude> </node> </tokenizerslangcontent>When building a Tokenizer, you can
customize its normalizer by just changing the corresponding attribute:
Of course, if you change the way a tokenizer applies normalization, you should probably retrain it from scratch afterward.
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
pre_tokenizers.Whitespace
pre-tokenizer:
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
offsets of our Encoding). Note that splitting on
punctuation will split contractions like "I'm" in this example.
You can combine together any PreTokenizer together. For instance, here is a pre-tokenizer that will
split on space, punctuation and digits, separating numbers in their
individual digits:
As we saw in the quicktour, you can
customize the pre-tokenizer of a Tokenizer by just changing the corresponding attribute:
Of course, if you change the way the pre-tokenizer, you should probably retrain your tokenizer from scratch afterward.
Once the input texts are normalized and pre-tokenized, the
Tokenizer applies 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 initializing the
Tokenizer so you already know how to
customize this part. Currently, the 🤗 Tokenizers library supports:
models.BPEmodels.Unigrammodels.WordLevelmodels.WordPieceFor more details about each model and its behavior, you can check here
Post-processing is the last step of the tokenization pipeline, to
perform any additional transformation to the
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
Tokenizer by setting the
corresponding attribute. For instance, here is how we can post-process
to make the inputs suitable for the BERT model:
Note that contrarily to the pre-tokenizer or the normalizer, you don't need to retrain a tokenizer after changing its post-processor.
Let's put all those pieces together to build a BERT tokenizer. First,
BERT relies on WordPiece, so we instantiate a new
Tokenizer with this model:
Then we know that BERT preprocesses texts by removing accents and lowercasing. We also use a unicode normalizer:
<tokenizerslangcontent> <python> <literalinclude> {"path": "../../bindings/python/tests/documentation/test_pipeline.py", "language": "python", "start-after": "START bert_setup_normalizer", "end-before": "END bert_setup_normalizer", "dedent": 8} </literalinclude> </python> <rust> <literalinclude> {"path": "../../tokenizers/tests/documentation.rs", "language": "rust", "start-after": "START bert_setup_normalizer", "end-before": "END bert_setup_normalizer", "dedent": 4} </literalinclude> </rust> <node> <literalinclude> {"path": "../../bindings/node/examples/documentation/pipeline.test.ts", "language": "js", "start-after": "START bert_setup_normalizer", "end-before": "END bert_setup_normalizer", "dedent": 8} </literalinclude> </node> </tokenizerslangcontent>The pre-tokenizer is just splitting on whitespace and punctuation:
<tokenizerslangcontent> <python> <literalinclude> {"path": "../../bindings/python/tests/documentation/test_pipeline.py", "language": "python", "start-after": "START bert_setup_pre_tokenizer", "end-before": "END bert_setup_pre_tokenizer", "dedent": 8} </literalinclude> </python> <rust> <literalinclude> {"path": "../../tokenizers/tests/documentation.rs", "language": "rust", "start-after": "START bert_setup_pre_tokenizer", "end-before": "END bert_setup_pre_tokenizer", "dedent": 4} </literalinclude> </rust> <node> <literalinclude> {"path": "../../bindings/node/examples/documentation/pipeline.test.ts", "language": "js", "start-after": "START bert_setup_pre_tokenizer", "end-before": "END bert_setup_pre_tokenizer", "dedent": 8} </literalinclude> </node> </tokenizerslangcontent>And the post-processing uses the template we saw in the previous section:
<tokenizerslangcontent> <python> <literalinclude> {"path": "../../bindings/python/tests/documentation/test_pipeline.py", "language": "python", "start-after": "START bert_setup_processor", "end-before": "END bert_setup_processor", "dedent": 8} </literalinclude> </python> <rust> <literalinclude> {"path": "../../tokenizers/tests/documentation.rs", "language": "rust", "start-after": "START bert_setup_processor", "end-before": "END bert_setup_processor", "dedent": 4} </literalinclude> </rust> <node> <literalinclude> {"path": "../../bindings/node/examples/documentation/pipeline.test.ts", "language": "js", "start-after": "START bert_setup_processor", "end-before": "END bert_setup_processor", "dedent": 8} </literalinclude> </node> </tokenizerslangcontent>We can use this tokenizer and train on it on wikitext like in the
quicktour:
On top of encoding the input texts, a 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
Tokenizer.decode (for one predicted text) and 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:
If you used a model that added special characters to represent subtokens
of a given "word" (like the "##" in
WordPiece) you will need to customize the decoder to treat
them properly. If we take our previous bert_tokenizer for instance the
default decoding will give:
But by changing it to a proper decoder, we get:
<tokenizerslangcontent> <python> <literalinclude> {"path": "../../bindings/python/tests/documentation/test_pipeline.py", "language": "python", "start-after": "START bert_proper_decoding", "end-before": "END bert_proper_decoding", "dedent": 8} </literalinclude> </python> <rust> <literalinclude> {"path": "../../tokenizers/tests/documentation.rs", "language": "rust", "start-after": "START bert_proper_decoding", "end-before": "END bert_proper_decoding", "dedent": 4} </literalinclude> </rust> <node> <literalinclude> {"path": "../../bindings/node/examples/documentation/pipeline.test.ts", "language": "js", "start-after": "START bert_proper_decoding", "end-before": "END bert_proper_decoding", "dedent": 8} </literalinclude> </node> </tokenizerslangcontent>