File size: 7,146 Bytes
6ae1fb5 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 | {
"cells": [
{
"cell_type": "markdown",
"id": "1f175efa",
"metadata": {},
"source": [
"# Latin Interpunctuator\n",
"\n",
"This notebook demonstrates how to use the mt5 model `mschonhardt/mt5-latin-punctuator-large`.\n",
"It applies interpunctuation and text formatting standards to Latin text.\n",
"\n",
"## Setup Environment"
]
},
{
"cell_type": "code",
"execution_count": 56,
"id": "044ae4ef",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Torch version: 2.10.0+cu128\n",
"Device: cuda\n",
"Environment ready.\n"
]
}
],
"source": [
"# Import necessary libraries\n",
"import torch\n",
"from transformers import AutoTokenizer, AutoModelForSeq2SeqLM\n",
"\n",
"# Model should be used with GPU (cuda) if available for faster inference\n",
"device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
"\n",
"print(f\"Torch version: {torch.__version__}\")\n",
"print(f\"Device: {device}\")\n",
"\n",
"print(\"Environment ready.\")"
]
},
{
"cell_type": "markdown",
"id": "4de2def2",
"metadata": {},
"source": [
"## Load the Model from Hugging Face"
]
},
{
"cell_type": "code",
"execution_count": 57,
"id": "aa5810a8",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading model: mschonhardt/mt5-latin-punctuator-large ...\n",
"Model loaded successfully!\n"
]
}
],
"source": [
"# Load the model and tokenizer from Huggingface\n",
"model_name = \"mschonhardt/mt5-latin-punctuator-large\" \n",
"print(f\"Loading model: {model_name} ...\")\n",
"tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)\n",
"model = AutoModelForSeq2SeqLM.from_pretrained(model_name).to(device)\n",
"print(\"Model loaded successfully!\")"
]
},
{
"cell_type": "markdown",
"id": "2dd05d72",
"metadata": {},
"source": [
"### Prediction Logic\n",
"Model was trained on prefix \"punctuate: \". `Num_beams` needs to be adjusted when running into hallucinations or repetitions. "
]
},
{
"cell_type": "code",
"execution_count": 59,
"id": "e858df99",
"metadata": {},
"outputs": [],
"source": [
"def punctuate(text: str) -> str:\n",
" # Best practice: Add prefix 'punctuate: 'and lowercase as per training script\n",
" input_text = \"punctuate: \" + text.lower()\n",
" \n",
" inputs = tokenizer(\n",
" input_text,\n",
" return_tensors=\"pt\",\n",
" truncation=True,\n",
" max_length=1024,\n",
" ).to(device)\n",
"\n",
" with torch.no_grad():\n",
" output_ids = model.generate(\n",
" **inputs,\n",
" max_length=1024,\n",
" # Adjust numbeams if hallucination occurs, but 4 is a good starting point for better punctuation\n",
" num_beams=4,\n",
" early_stopping=True,\n",
" )\n",
" return tokenizer.decode(output_ids[0], skip_special_tokens=True)\n"
]
},
{
"cell_type": "code",
"execution_count": 60,
"id": "52fd09e1",
"metadata": {},
"outputs": [],
"source": [
"text = \"\"\"\n",
"Si quis Patrem et Filium et Spiritum Sanctum non confitetur tres personas unius substantiae et virtutis ac potestatis, \n",
"sicut catholica et apostolica ecclesia docet, sed unam tantum ac solitariam dicit esse personam, \n",
"ita ut ipse sit Pater qui Filius, ipse etiam sit Paraclitus Spiritus, sicut Sabellius et Priscillianus dixerunt, anathema sit.\"\"\""
]
},
{
"cell_type": "markdown",
"id": "e582b0e4",
"metadata": {},
"source": [
"Model was trained on lower case input to prevent overfitting on capital letters and force learning of linguistic pattern."
]
},
{
"cell_type": "code",
"execution_count": 61,
"id": "6573900a",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
" si quis patrem et filium et spiritum sanctum non confitetur tres personas unius\n",
"substantiae et virtutis ac potestatis sicut catholica et apostolica ecclesia\n",
"docet sed unam tantum ac solitariam dicit esse personam ita ut ipse sit pater\n",
"qui filius ipse etiam sit paraclitus spiritus sicut sabellius et priscillianus\n",
"dixerunt anathema sit\n"
]
}
],
"source": [
"text_without_punctuation = text.replace(\".\",\"\").replace(\",\",\"\").replace(\";\",\"\").replace(\":\",\"\").replace(\"?\",\"\").replace(\"!\",\"\").replace(\"-\",\"\").replace(\"(\",\"\").replace(\")\",\"\").replace(\"[\",\"\").replace(\"]\",\"\").replace(\"{\",\"\").replace(\"}\",\"\").replace(\"\\\"\",\"\")\n",
"text_without_punctuation = text_without_punctuation.lower()\n",
"import textwrap\n",
"print(textwrap.fill(text_without_punctuation, width=80))"
]
},
{
"cell_type": "markdown",
"id": "843757c0",
"metadata": {},
"source": [
"### Run Inference"
]
},
{
"cell_type": "code",
"execution_count": 65,
"id": "86c7521d",
"metadata": {},
"outputs": [],
"source": [
"# Model will predict punctuation for the input text as well as appropriate use of capital letters\n",
"# Note: The model will reflect conventions of material it has seen, which might differ from your expectations.\n",
"text_with_punctuation = punctuate(text_without_punctuation)\n"
]
},
{
"cell_type": "markdown",
"id": "02ef7e70",
"metadata": {},
"source": [
"As the model does apply conventions it has learned from training data, the models decision might differ from your own conventions and expectations. It has not been designed to prepare a 'perfect' text, but to provide structure to unstrucutred text enabling downstream tasks,"
]
},
{
"cell_type": "code",
"execution_count": 66,
"id": "1c908d35",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Si quis Patrem et Filium et Spiritum sanctum non confitetur tres personas unius\n",
"substantiae et virtutis ac potestatis, sicut catholica et apostolica Ecclesia\n",
"docet, sed unam tantum ac solitariam dicit esse personam, ita ut ipse sit Pater\n",
"qui Filius, ipse etiam sit Paraclitus Spiritus, sicut Sabellius et Priscillianus\n",
"dixerunt, anathema sit.\n"
]
}
],
"source": [
"print(textwrap.fill(text_with_punctuation, width=80))"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "venv-jupyter",
"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": "3.12.3"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|