Tokenizers documentation


You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v0.13.4.rc2).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started




class tokenizers.pre_tokenizers.BertPreTokenizer

( )


This pre-tokenizer splits tokens on spaces, and also on punctuation. Each occurence of a punctuation character will be treated separately.


class tokenizers.pre_tokenizers.ByteLevel

( add_prefix_space = True use_regex = True )


  • add_prefix_space (bool, optional, defaults to True) — Whether to add a space to the first word if there isn’t already one. This lets us treat hello exactly like say hello.
  • use_regex (bool, optional, defaults to True) — Set this to False to prevent this pre_tokenizer from using the GPT2 specific regexp for spliting on whitespace.

ByteLevel PreTokenizer

This pre-tokenizer takes care of replacing all bytes of the given string with a corresponding representation, as well as splitting into words.


( ) β†’ List[str]



A list of characters that compose the alphabet

Returns the alphabet used by this PreTokenizer.

Since the ByteLevel works as its name suggests, at the byte level, it encodes each byte value to a unique visible character. This means that there is a total of 256 different characters composing this alphabet.


class tokenizers.pre_tokenizers.CharDelimiterSplit

( )

This pre-tokenizer simply splits on the provided char. Works like .split(delimiter)


class tokenizers.pre_tokenizers.Digits

( individual_digits = False )


  • individual_digits (bool, optional, defaults to False) —

This pre-tokenizer simply splits using the digits in separate tokens

If set to True, digits will each be separated as follows:

"Call 123 please" -> "Call ", "1", "2", "3", " please"

If set to False, digits will grouped as follows:

"Call 123 please" -> "Call ", "123", " please"


class tokenizers.pre_tokenizers.Metaspace

( replacement = '_' prepend_scheme = 'always' split = True )


  • replacement (str, optional, defaults to ) — The replacement character. Must be exactly one character. By default we use the (U+2581) meta symbol (Same as in SentencePiece).
  • prepend_scheme (str, optional, defaults to "always") — Whether to add a space to the first word if there isn’t already one. This lets us treat hello exactly like say hello. Choices: “always”, “never”, “first”. First means the space is only added on the first token (relevant when special tokens are used or other pre_tokenizer are used).

Metaspace pre-tokenizer

This pre-tokenizer replaces any whitespace by the provided replacement character. It then tries to split on these spaces.


class tokenizers.pre_tokenizers.PreTokenizer

( )

Base class for all pre-tokenizers

This class is not supposed to be instantiated directly. Instead, any implementation of a PreTokenizer will return an instance of this class when instantiated.


( pretok )


  • pretok (~tokenizers.PreTokenizedString) -- The pre-tokenized string on which to apply this :class:~tokenizers.pre_tokenizers.PreTokenizer`

Pre-tokenize a ~tokenizers.PyPreTokenizedString in-place

This method allows to modify a PreTokenizedString to keep track of the pre-tokenization, and leverage the capabilities of the PreTokenizedString. If you just want to see the result of the pre-tokenization of a raw string, you can use pre_tokenize_str()


( sequence ) β†’ List[Tuple[str, Offsets]]


  • sequence (str) — A string to pre-tokeize


List[Tuple[str, Offsets]]

A list of tuple with the pre-tokenized parts and their offsets

Pre tokenize the given string

This method provides a way to visualize the effect of a PreTokenizer but it does not keep track of the alignment, nor does it provide all the capabilities of the PreTokenizedString. If you need some of these, you can use pre_tokenize()

## Punctuation[[tokenizers.pre_tokenizers.Punctuation]]

class tokenizers.pre_tokenizers.Punctuation

( behavior = 'isolated' )


  • behavior (SplitDelimiterBehavior) — The behavior to use when splitting. Choices: “removed”, “isolated” (default), “merged_with_previous”, “merged_with_next”, “contiguous”

This pre-tokenizer simply splits on punctuation as individual characters.

## Sequence[[tokenizers.pre_tokenizers.Sequence]]

class tokenizers.pre_tokenizers.Sequence

( pretokenizers )

This pre-tokenizer composes other pre_tokenizers and applies them in sequence

## Split[[tokenizers.pre_tokenizers.Split]]

class tokenizers.pre_tokenizers.Split

( pattern behavior invert = False )


  • pattern (str or Regex) — A pattern used to split the string. Usually a string or a a regex built with tokenizers.Regex
  • behavior (SplitDelimiterBehavior) — The behavior to use when splitting. Choices: “removed”, “isolated”, “merged_with_previous”, “merged_with_next”, “contiguous”
  • invert (bool, optional, defaults to False) — Whether to invert the pattern.

Split PreTokenizer

This versatile pre-tokenizer splits using the provided pattern and according to the provided behavior. The pattern can be inverted by making use of the invert flag.

## UnicodeScripts[[tokenizers.pre_tokenizers.UnicodeScripts]]

class tokenizers.pre_tokenizers.UnicodeScripts

( )

This pre-tokenizer splits on characters that belong to different language family It roughly follows Actually Hiragana and Katakana are fused with Han, and 0x30FC is Han too. This mimicks SentencePiece Unigram implementation.

## Whitespace[[tokenizers.pre_tokenizers.Whitespace]]

class tokenizers.pre_tokenizers.Whitespace

( )

This pre-tokenizer simply splits using the following regex: \w+|[^\w\s]+

## WhitespaceSplit[[tokenizers.pre_tokenizers.WhitespaceSplit]]

class tokenizers.pre_tokenizers.WhitespaceSplit

( )

This pre-tokenizer simply splits on the whitespace. Works like .split()

< > Update on GitHub