Tokenizer base class#

class lmp.tknzr.BaseTknzr(*, is_uncased: bool = False, max_vocab: int = -1, min_count: int = 0, **kwargs: Any)[source]#

Bases: ABC

Tokenizer abstract base class.

Provide text processing functionalities including tokenization, normalization and language model training formation.

This class is designed to be the abstract base class of all tokenizers. Both tokenization and detokenization are left unimplemented.

Parameters

is_uncased (bool, default: False) – Set to True to convert text into lowercase. Mainly used by norm.
max_vocab (int, default: -1) – Tokenizer’s maximum vocabulary size. Set to -1 to include as many tokens in vocabulary as possible. Mainly used by build_vocab.
min_count (int, default: 0) – Minimum token occurrence counts. Tokens have occurrence counts less than min_count will not be added to tokenizer’s vocabulary. Mainly used by build_vocab.
kwargs (Any, optional) – Useless parameter. Intently left for subclasses inheritance.

id2tk#

Token-to-id inverse lookup table.

Type: dict[int, str]

is_uncased#

Convert text into lowercase if set to True.

Type: bool

max_vocab#

Tokenizer’s maximum vocabulary size.

Type: int

min_count#

Minimum token occurrence counts.

Type: int

tk2id#

Token-to-id lookup table.

Type: dict[str, int]

tknzr_name#

CLI name of the tokenizer. Only used to parse CLI arguments.

Type: ClassVar[str]

See also

lmp.tknzr: All available tokenizers.

classmethod add_CLI_args(parser: ArgumentParser) → None[source]#

Add tokenizer hyperparameters to CLI argument parser.

Parameters: parser (argparse.ArgumentParser) – CLI argument parser.
Return type: None

See also

lmp.script.train_tknzr: Tokenizer training script.

Examples

>>> import argparse
>>> from lmp.tknzr import BaseTknzr
>>> parser = argparse.ArgumentParser()
>>> BaseTknzr.add_CLI_args(parser)
>>> args = parser.parse_args([
...   '--max_vocab', '10',
...   '--min_count', '2',
... ])
>>> assert args.is_uncased == False
>>> assert args.max_vocab == 10
>>> assert args.min_count == 2

build_vocab(batch_txt: Iterable[str]) → None[source]#

Build tokenizer’s vocabulary.

Build vocabulary based on token occurrence counts. Text in batch_txt is first normalized and tokenized, then count each token’s occurrence. Tokens with higher occurrence counts are added to vocabulary first. Tokens with the same occurrence counts are added to vocabulary in the order of their appearance.

When adding a new token to vocabulary, its token id will be assign to the largest token id + 1. Tokens already in vocabulary are not added to vocabulary again. If a token’s occurrence count is lower than self.min_count, then that token is not added to vocabulary. If vocabulary size is larger than or equal to self.max_vocab, then no new tokens are added to vocabulary.

Parameters: batch_txt (collections.abc.Iterable[str]) – Source of text to build vocabulary.
Return type: None

See also

norm: Perform normalization on text.
tknz: Perform tokenization on text.
vocab_size: Tokenizer’s vocabulary size.

dec(tkids: List[int], *, rm_sp_tks: bool = False) → str[source]#

Decode token id list back to text.

Token id list is first converted into token list then detokenized back to text. Special tokens other than <unk> will be removed if setting rm_sp_tks=True. Token ids not in tokenizer’s inverse lookup table are converted into <unk> token.

Parameters

tkids (list[int]) – Token id list to be decoded.
rm_sp_tks (bool, default: False) – Set to True to remove <bos>, <eos> and <pad>.

Returns

Decoded text.

Return type

str

See also

dtknz: Convert tokens back to text.
enc: Encode text into token id list.

Note

Unknown tokens <unk> will not be removed even if setting rm_sp_tks=True. This is simply because we do not know which token to convert it back (thus the name unknown token).

abstract dtknz(tks: List[str]) → str[source]#

Convert tokens back to text.

Tokens will be detokenized and normalized by norm. The execution order of detokenization and normalization will not effect the result.

Parameters: tks (list[str]) – Token list to be detokenized.
Returns: Text which is normalized and detokenized from token list.
Return type: str

See also

tknz: Tokenize text into token list.
norm: Text normalization.

enc(txt: str) → List[int][source]#

Encode text into token id list.

Text will be tokenized into token list (tk_0, tk_1, ..., tk_n) and formatted as follow:

<bos> tk_0 tk_1 ... tk_n <eos>

<bos> is the “begin of sequence” token.
<eos> is the “end of sequence” token.
<unk> token is used to replace OOV tokens.

All tokens in token list are converted into token ids and returned.

Parameters: txt (str) – Text to be encoded.
Returns: Token ids list.
Return type: list[int]

See also

dec: Decode token id list back to text.
pad_to_max: Pad token id list to specified length.
tknz: Perform tokenization on text.

norm(txt: str) → str[source]#

Perform text normalization.

Text are normalized by NFKC. Whitespaces are collapsed and stripped from both ends. Text are converted into lowercase if setting is_uncased=True.

Parameters: txt (str) – Text to be normalized.
Returns: Normalized text.
Return type: str

See also

unicodedata.normalize: Python built-in unicode normalization.

Examples

Convert text to lowercase.

>>> from lmp.tknzr import CharTknzr
>>> tknzr = CharTknzr(is_uncased=True)
>>> assert tknzr.norm('ABC') == 'abc'

pad_to_max(max_seq_len: int, tkids: List[int]) → List[int][source]#

Pad token id list to specified length.

If len(tkids) < max_seq_len, then append padding token id at the end of tkids until tkids has length equal to max_seq_len. Do nothing when len(tkids) >= max_seq_len.

Parameters

max_seq_len (int) – Maximum length constraint.
tkids (list[int]) – Token id list to be padded.

Returns

Padded token id list.

Return type

list[int]

Examples

>>> from lmp.vars import PAD_TKID
>>> from lmp.tknzr import CharTknzr
>>> tknzr = CharTknzr()
>>> assert tknzr.pad_to_max(max_seq_len=4, tkids=[1, 2, 3]) == [1, 2, 3, PAD_TKID]

abstract tknz(txt: str) → List[str][source]#

Perform tokenization on text.

Text is first normalized then tokenized into token list.

Parameters: txt (str) – Text to be tokenized.
Returns: List of normalized tokens.
Return type: list[str]

See also

dtknz: Detokenize token list back to text.
norm: Text normalization.

property vocab_size: int#

Get tokenizer vocabulary size.

Returns: Tokenizer vocabulary size.
Return type: int

See also

build_vocab: Build vocabulary for tokenizer.