splintr 0.12.0

Fast Rust tokenizer (BPE + SentencePiece + WordPiece) with 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

//! Errors from loading a HuggingFace `tokenizer.json`.

use thiserror::Error;

use super::super::policy::PolicyError;
use super::super::sentencepiece::SentencePieceError;
use super::super::tokenizer::TokenizerError;
use super::super::wordpiece::WordPieceError;

/// Errors from loading a HuggingFace `tokenizer.json`.
#[derive(Debug, Error)]
pub enum HfJsonError {
    #[error("failed to parse tokenizer.json: {0}")]
    Json(#[from] serde_json::Error),
    #[error("tokenizer.json missing field: {0}")]
    MissingField(&'static str),
    #[error("unsupported model.type `{0}` (expected BPE, Unigram, or WordPiece)")]
    UnsupportedModelType(String),
    #[error(
        "unsupported normalizer type(s) `{0}` — refusing to load rather than silently drop them"
    )]
    UnsupportedNormalizer(String),
    #[error("normalizer Replace pattern `{0}` failed to compile as a regex")]
    InvalidNormalizerRegex(String),
    #[error("unsupported pre_tokenizer type(s) `{0}` and no recognized split — refusing to guess the split pattern")]
    UnsupportedPreTokenizer(String),
    #[error("vocab entry `{0}` is not valid byte-level encoding")]
    InvalidByteLevel(String),
    #[error(
        "added token `{content}` is listed in model.vocab with id {vocab_id} but in added_tokens with id {added_id}"
    )]
    AddedTokenIdConflict {
        /// The token spelling both sections declare.
        content: String,
        /// The id `model.vocab` gives it (what BPE/decode would use).
        vocab_id: u32,
        /// The id `added_tokens` gives it (what the matcher would emit).
        added_id: u32,
    },
    #[error("could not determine the {0} token id from the tokenizer.json")]
    MissingSpecial(&'static str),
    #[error(transparent)]
    Policy(#[from] PolicyError),
    #[error(transparent)]
    Tokenizer(#[from] TokenizerError),
    #[error(transparent)]
    SentencePiece(#[from] SentencePieceError),
    #[error(transparent)]
    WordPiece(#[from] WordPieceError),
    #[error("I/O error: {0}")]
    Io(#[from] std::io::Error),
}