File size: 6,936 Bytes
2d923bf |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 |
# Copyright (C) Miðeind ehf.
# This file is part of IceBERT POS model conversion.
import json
from typing import Any, Dict, Optional
from transformers import AutoConfig, RobertaConfig
class IceBertPosConfig(RobertaConfig):
"""
Configuration class for IceBERT POS (Part-of-Speech) tagging model.
This configuration inherits from RobertaConfig and adds POS-specific parameters
derived from the label schema used for multilabel token classification.
"""
model_type = "icebert-pos"
def __init__(
self, label_schema: Optional[Dict[str, Any]] = None, classifier_dropout: Optional[float] = None, **kwargs
):
super().__init__(**kwargs)
# Default label schema (terms2.json content)
if label_schema is None:
label_schema = self._get_default_label_schema()
self.label_schema = label_schema
# Derive parameters from label schema
self.num_categories = len(label_schema["label_categories"])
self.num_labels = len(label_schema["labels"])
self.num_groups = len(label_schema["group_names"])
# Classification head parameters
self.classifier_dropout = classifier_dropout if classifier_dropout is not None else 0.1
# Computed input size for attribute projection
# (category_probs + hidden_size) -> num_labels
self.attr_proj_input_size = self.num_categories + self.hidden_size
@staticmethod
def _get_default_label_schema() -> Dict[str, Any]:
"""Default label schema corresponding to terms2.json"""
return {
"label_categories": [
"n",
"g",
"x",
"e",
"v",
"l",
"fa",
"fb",
"fe",
"fo",
"fp",
"fs",
"ft",
"tf",
"ta",
"tp",
"to",
"sn",
"sb",
"sf",
"sv",
"ss",
"sl",
"sþ",
"cn",
"ct",
"c",
"aa",
"af",
"au",
"ao",
"aþ",
"ae",
"as",
"ks",
"kt",
"p",
"pl",
"pk",
"pg",
"pa",
"ns",
"m",
],
"category_to_group_names": {
"n": ["gender", "number", "case", "def", "proper"],
"g": ["gender", "number", "case"],
"l": ["gender", "number", "case", "adj_c", "deg"],
"fa": ["gender", "number", "case"],
"fb": ["gender", "number", "case"],
"fe": ["gender", "number", "case"],
"fs": ["gender", "number", "case"],
"ft": ["gender", "number", "case"],
"fo": ["gender_or_person", "number", "case"],
"fp": ["gender_or_person", "number", "case"],
"tf": ["gender", "number", "case"],
"sn": ["voice"],
"sb": ["voice", "person", "number", "tense"],
"sf": ["voice", "person", "number", "tense"],
"sv": ["voice", "person", "number", "tense"],
"ss": ["voice"],
"sl": ["voice", "person", "number", "tense"],
"sþ": ["voice", "gender", "number", "case"],
"aa": ["deg"],
"af": ["deg"],
"au": ["deg"],
"ao": ["deg"],
"aþ": ["deg"],
"ae": ["deg"],
"as": ["deg"],
},
"group_names": [
"gender",
"gender_or_person",
"number",
"case",
"def",
"proper",
"adj_c",
"deg",
"voice",
"person",
"tense",
],
"group_name_to_labels": {
"gender": ["masc", "fem", "neut", "gender_x"],
"number": ["sing", "plur"],
"person": ["1", "2", "3"],
"gender_or_person": ["masc", "fem", "neut", "gender_x", "1", "2", "3"],
"case": ["nom", "acc", "dat", "gen"],
"deg": ["pos", "cmp", "superl"],
"voice": ["act", "mid"],
"tense": ["pres", "past"],
"def": ["definite"],
"proper": ["proper"],
"adj_c": ["strong", "weak", "equiinflected"],
},
"labels": [
"<SEP>",
"n",
"g",
"x",
"e",
"v",
"l",
"fa",
"fb",
"fe",
"fo",
"fp",
"fs",
"ft",
"tf",
"ta",
"tp",
"to",
"sn",
"sb",
"sf",
"sv",
"ss",
"sl",
"sþ",
"cn",
"ct",
"c",
"aa",
"af",
"au",
"ao",
"aþ",
"ae",
"as",
"ks",
"kt",
"p",
"pl",
"pk",
"pg",
"pa",
"ns",
"m",
"masc",
"fem",
"neut",
"gender_x",
"1",
"2",
"3",
"sing",
"plur",
"nom",
"acc",
"dat",
"gen",
"definite",
"proper",
"strong",
"weak",
"equiinflected",
"pos",
"cmp",
"superl",
"past",
"pres",
"pass",
"act",
"mid",
],
"null": None,
"null_leaf": None,
"separator": "<SEP>",
"ignore_categories": ["x", "e"],
}
@classmethod
def from_label_schema_file(cls, schema_path: str, **kwargs) -> "IceBertPosConfig":
"""Create config from a label schema JSON file"""
with open(schema_path, "r", encoding="utf-8") as f:
label_schema = json.load(f)
return cls(label_schema=label_schema, **kwargs)
AutoConfig.register("icebert-pos", IceBertPosConfig)
|