File size: 5,061 Bytes
e5aaad7 |
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 |
{
"cells": [
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"# Now training Wordpiece with instructions from huggingface\n"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"# https://colab.research.google.com/github/huggingface/notebooks/blob/master/examples/tokenizer_training.ipynb\n"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"# join all txt files into a single txt file\n",
"import os\n",
"from pathlib import Path\n",
"import time\n",
"\n",
"paths = [str(x) for x in Path(\"./custom_latin_corpus\").glob(\"**/*.txt\")]\n",
"all_text = []\n",
"for path in paths:\n",
" with open(path, \"r\") as f:\n",
" text = f.read()\n",
"\n",
" all_text.append(text)\n",
"# text batch size\n",
"batch_size = 100\n",
"def batch_iterator():\n",
" for i in range(0, len(all_text), batch_size):\n",
" yield all_text[i : i + batch_size]"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"with open(\"03_full_latin_corpus_for_training.txt\", \"w\") as f:\n",
" f.writelines(all_text)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"\n",
"from tokenizers import decoders, models, normalizers, pre_tokenizers, processors, trainers, Tokenizer\n"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"tokenizer = Tokenizer(models.WordPiece(unk_token=\"[UNK]\"))\n",
"tokenizer.normalizer = normalizers.BertNormalizer(lowercase=True)\n",
"tokenizer.pre_tokenizer = pre_tokenizers.BertPreTokenizer()\n",
"tokenizer.pre_tokenizer.pre_tokenize_str(\"This is an example!\")\n",
"special_tokens = [\"[UNK]\", \"[PAD]\", \"[CLS]\", \"[SEP]\", \"[MASK]\"]"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"\n",
"trainer = trainers.WordPieceTrainer(\n",
" vocab_size=25000, \n",
" special_tokens=special_tokens,\n",
" min_frequency=2,\n",
" limit_alphabet=50\n",
" )"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"\n",
"\n"
]
}
],
"source": [
"\n",
"\n",
"\n",
"tokenizer.train_from_iterator(batch_iterator(), trainer=trainer)\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"2 3\n"
]
}
],
"source": [
"# now, define the post processor\n",
"cls_token_id = tokenizer.token_to_id(\"[CLS]\")\n",
"sep_token_id = tokenizer.token_to_id(\"[SEP]\")\n",
"print(cls_token_id, sep_token_id)\n",
"tokenizer.post_processor = processors.TemplateProcessing(\n",
" single=f\"[CLS]:0 $A:0 [SEP]:0\",\n",
" pair=f\"[CLS]:0 $A:0 [SEP]:0 $B:1 [SEP]:1\",\n",
" special_tokens=[\n",
" (\"[CLS]\", cls_token_id),\n",
" (\"[SEP]\", sep_token_id),\n",
" ],\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"# test an example\n",
"example_encoding = tokenizer.encode(\"Roma in Italia est.\", \"Italia in Europa est.\")\n",
"example_encoding.tokens\n",
"tokenizer.decoder = decoders.WordPiece(prefix=\"##\")\n"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"('latin_WP_tokenizer/tokenizer_config.json',\n",
" 'latin_WP_tokenizer/special_tokens_map.json',\n",
" 'latin_WP_tokenizer/vocab.txt',\n",
" 'latin_WP_tokenizer/added_tokens.json',\n",
" 'latin_WP_tokenizer/tokenizer.json')"
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# wrap it inside transformers object\n",
"\n",
"from transformers import BertTokenizerFast\n",
"\n",
"new_wp_tokenizer = BertTokenizerFast(tokenizer_object=tokenizer)\n",
"new_wp_tokenizer.save_pretrained(\n",
" \"latin_WP_tokenizer\"\n",
")\n"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "bertenv",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.2"
},
"orig_nbformat": 4
},
"nbformat": 4,
"nbformat_minor": 2
}
|