unitoken 0.1.4

Fast BPE tokenizer/trainer with a Rust core and Python bindings
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66

from __future__ import annotations

from collections.abc import Sequence
from os import PathLike
from typing import Literal

from ._lib import PreTokenizer as _PreTokenizer


BoundaryMode = Literal["auto", "eot", "line", "utf8"]
UnicodeBigramMixedBoundary = Literal["keep", "split"]


class PreTokenizer:
  def __init__(
    self,
    special_tokens: Sequence[str],
    eot_token: str | None = None,
    pat: str | None = None,
    unicode_bigrams: Sequence[str] | None = None,
    unicode_bigram_mixed_boundary: UnicodeBigramMixedBoundary = "keep",
  ) -> None:
    self._inner = _PreTokenizer(special_tokens, eot_token, pat, unicode_bigrams, unicode_bigram_mixed_boundary)

  def find_chunk_boundaries(
    self,
    path: str | PathLike,
    *,
    chunk_size: int = 1024 * 1024,
    boundary: BoundaryMode = "auto",
  ) -> list[tuple[int, int]]:
    return self._inner.find_chunk_boundaries(path, chunk_size=chunk_size, boundary=boundary)

  def get_words_from_file(
    self,
    path: str | PathLike,
    *,
    chunk_size: int = 1024 * 1024,
    boundary: BoundaryMode = "auto",
  ) -> dict[str, int]:
    return self._inner.get_words_from_file(path, chunk_size=chunk_size, boundary=boundary)

  def get_words_from_segment(
    self,
    path: str | PathLike,
    offset: int,
    length: int,
  ) -> dict[str, int]:
    return self._inner.get_words_from_segment(path, offset, length)

  def build_unicode_bigrams_from_file(
    self,
    path: str | PathLike,
    *,
    chunk_size: int = 1024 * 1024,
    boundary: BoundaryMode = "auto",
    top_k: int = 100_000,
    min_freq: int = 16,
  ) -> list[str]:
    return self._inner.build_unicode_bigrams_from_file(
      path,
      chunk_size=chunk_size,
      boundary=boundary,
      top_k=top_k,
      min_freq=min_freq,
    )