Hugging Face's logo

janakhpon
/
mon_tokenizer

Mon
Burmese
English
tokenizers
tokenizer
unigram
mon
burmese
myanmar
low-resource
Model card Files
xet
 Community
mon_tokenizer
File size: 2,717 Bytes
954bd9d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b2eb67d
 
 
954bd9d
 
 
 
 
 
 
 
 
 
b2eb67d
 
954bd9d
 
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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
{
  "name": "mon-tokenizer",
  "artifact_version": "1.0.0",
  "algorithm": "unigram",
  "vocab_size": 64256,
  "languages": [
    "mnw",
    "mya",
    "eng"
  ],
  "config": {
    "vocab_size": 64000,
    "special_tokens": [
      "<unk>",
      "<s>",
      "</s>",
      "<pad>"
    ],
    "unk_token": "<unk>",
    "byte_fallback": true,
    "enforce_syllable_atomicity": false,
    "max_piece_length": 16
  },
  "corpus": {
    "lines": 893936,
    "chars": 85793599,
    "by_bucket": {
      "mon": 592670,
      "burmese": 36341,
      "english": 264925
    },
    "source_digest": "11941a573a5e0c618edbc91d34dd787d"
  },
  "metrics": {
    "mon": {
      "lines": 29600,
      "chars": 2280192,
      "tokens": 486631,
      "chars_per_token": 4.686,
      "tokens_per_line": 16.4,
      "roundtrip": 1.0,
      "fallback_rate": 4e-05,
      "violation_rate": 0.01074,
      "syllables": 492469,
      "unreconstructable": 4
    },
    "burmese": {
      "lines": 1792,
      "chars": 102682,
      "tokens": 24942,
      "chars_per_token": 4.117,
      "tokens_per_line": 13.9,
      "roundtrip": 1.0,
      "fallback_rate": 0.0,
      "violation_rate": 0.00928,
      "syllables": 25546,
      "unreconstructable": 0
    },
    "english": {
      "lines": 13104,
      "chars": 1864318,
      "tokens": 453350,
      "chars_per_token": 4.112,
      "tokens_per_line": 34.6,
      "roundtrip": 1.0,
      "fallback_rate": 0.00021,
      "violation_rate": 0.0,
      "syllables": 0,
      "unreconstructable": 7
    },
    "mixed": {
      "lines": 2420,
      "chars": 226127,
      "tokens": 59440,
      "chars_per_token": 3.804,
      "tokens_per_line": 24.6,
      "roundtrip": 1.0,
      "fallback_rate": 0.00195,
      "violation_rate": 0.0081,
      "syllables": 28133,
      "unreconstructable": 11
    },
    "coverage": {
      "distinct": 397,
      "single_token": 0.9874,
      "lines_measured": 29600
    }
  },
  "notes": {
    "compression_is_corpus_dependent": "chars_per_token describes this corpus, not the tokenizer. Quote it with the corpus or not at all.",
    "violation_rate": "Share of multi-character Myanmar syllables split by a token boundary, using the segmenter in syllable.py rather than Unicode grapheme clusters, which do not bound a Myanmar syllable. Report the denominator with the rate.",
    "roundtrip": "Compared after normalization on both sides. Normalization is a deliberate transform, not a loss.",
    "byte_fallback": "Every input round-trips, including Thai, emoji and CJK, at 1-4 tokens per character. Only the guaranteed alphabet is cheap."
  },
  "eval": {
    "split": "val",
    "sampling": "whole split, no cap",
    "coverage_lines_measured": 29600
  }
}