diff --git a/klue_re.ipynb b/klue_re.ipynb
index 389e40e..a0d909f 100644
--- a/klue_re.ipynb
+++ b/klue_re.ipynb
@@ -2,7 +2,7 @@
"cells": [
{
"cell_type": "markdown",
- "id": "impressive-sociology",
+ "id": "concerned-regulation",
"metadata": {},
"source": [
"# BERT를 활용한 관계추출(Relation Extraction, RE)\n",
@@ -29,7 +29,7 @@
},
{
"cell_type": "markdown",
- "id": "august-garden",
+ "id": "lightweight-application",
"metadata": {},
"source": [
"# 필요한 라이브러리를 설치합니다.\n",
@@ -39,43 +39,43 @@
{
"cell_type": "code",
"execution_count": 1,
- "id": "collective-number",
+ "id": "inner-crime",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "Requirement already satisfied: datasets in /opt/conda/lib/python3.7/site-packages (1.8.0)\n",
- "Requirement already satisfied: huggingface-hub<0.1.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (0.0.13)\n",
- "Requirement already satisfied: requests>=2.19.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (2.25.1)\n",
- "Requirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from datasets) (3.7.3)\n",
- "Requirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (from datasets) (1.2.5)\n",
- "Requirement already satisfied: multiprocess in /opt/conda/lib/python3.7/site-packages (from datasets) (0.70.12.2)\n",
- "Requirement already satisfied: numpy>=1.17 in /opt/conda/lib/python3.7/site-packages (from datasets) (1.18.5)\n",
- "Requirement already satisfied: pyarrow<4.0.0,>=1.0.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (3.0.0)\n",
- "Requirement already satisfied: tqdm<4.50.0,>=4.27 in /opt/conda/lib/python3.7/site-packages (from datasets) (4.49.0)\n",
- "Requirement already satisfied: xxhash in /opt/conda/lib/python3.7/site-packages (from datasets) (2.0.2)\n",
- "Requirement already satisfied: fsspec in /opt/conda/lib/python3.7/site-packages (from datasets) (2021.6.1)\n",
+ "Requirement already satisfied: datasets in /opt/conda/lib/python3.7/site-packages (1.9.0)\n",
+ "Requirement already satisfied: tqdm>=4.27 in /opt/conda/lib/python3.7/site-packages (from datasets) (4.59.0)\n",
"Requirement already satisfied: packaging in /opt/conda/lib/python3.7/site-packages (from datasets) (20.9)\n",
+ "Requirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (from datasets) (1.3.0)\n",
"Requirement already satisfied: dill in /opt/conda/lib/python3.7/site-packages (from datasets) (0.3.4)\n",
- "Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from huggingface-hub<0.1.0->datasets) (3.7.4.3)\n",
+ "Requirement already satisfied: xxhash in /opt/conda/lib/python3.7/site-packages (from datasets) (2.0.2)\n",
+ "Requirement already satisfied: huggingface-hub<0.1.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (0.0.13)\n",
+ "Requirement already satisfied: pyarrow!=4.0.0,>=1.0.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (4.0.1)\n",
+ "Requirement already satisfied: numpy>=1.17 in /opt/conda/lib/python3.7/site-packages (from datasets) (1.18.5)\n",
+ "Requirement already satisfied: multiprocess in /opt/conda/lib/python3.7/site-packages (from datasets) (0.70.12.2)\n",
+ "Requirement already satisfied: fsspec>=2021.05.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (2021.6.1)\n",
+ "Requirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from datasets) (3.7.3)\n",
+ "Requirement already satisfied: requests>=2.19.0 in /opt/conda/lib/python3.7/site-packages (from datasets) (2.25.1)\n",
"Requirement already satisfied: filelock in /opt/conda/lib/python3.7/site-packages (from huggingface-hub<0.1.0->datasets) (3.0.12)\n",
+ "Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from huggingface-hub<0.1.0->datasets) (3.7.4.3)\n",
"Requirement already satisfied: pyparsing>=2.0.2 in /opt/conda/lib/python3.7/site-packages (from packaging->datasets) (2.4.7)\n",
- "Requirement already satisfied: chardet<5,>=3.0.2 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (4.0.0)\n",
- "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (2020.12.5)\n",
"Requirement already satisfied: idna<3,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (2.10)\n",
+ "Requirement already satisfied: chardet<5,>=3.0.2 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (4.0.0)\n",
"Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (1.26.4)\n",
+ "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests>=2.19.0->datasets) (2020.12.5)\n",
"Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->datasets) (3.4.1)\n",
- "Requirement already satisfied: python-dateutil>=2.7.3 in /opt/conda/lib/python3.7/site-packages (from pandas->datasets) (2.8.1)\n",
"Requirement already satisfied: pytz>=2017.3 in /opt/conda/lib/python3.7/site-packages (from pandas->datasets) (2021.1)\n",
+ "Requirement already satisfied: python-dateutil>=2.7.3 in /opt/conda/lib/python3.7/site-packages (from pandas->datasets) (2.8.1)\n",
"Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7.3->pandas->datasets) (1.15.0)\n",
"Requirement already satisfied: sklearn in /opt/conda/lib/python3.7/site-packages (0.0)\n",
"Requirement already satisfied: scikit-learn in /opt/conda/lib/python3.7/site-packages (from sklearn) (0.24.2)\n",
- "Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (2.1.0)\n",
- "Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (1.0.1)\n",
"Requirement already satisfied: scipy>=0.19.1 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (1.4.1)\n",
- "Requirement already satisfied: numpy>=1.13.3 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (1.18.5)\n"
+ "Requirement already satisfied: numpy>=1.13.3 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (1.18.5)\n",
+ "Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (2.1.0)\n",
+ "Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->sklearn) (1.0.1)\n"
]
}
],
@@ -86,7 +86,7 @@
},
{
"cell_type": "markdown",
- "id": "organic-valley",
+ "id": "critical-trading",
"metadata": {},
"source": [
"# 필요한 라이브러리를 import 합니다."
@@ -95,7 +95,7 @@
{
"cell_type": "code",
"execution_count": 2,
- "id": "looking-relevance",
+ "id": "regulated-musician",
"metadata": {},
"outputs": [],
"source": [
@@ -106,14 +106,13 @@
"from tqdm import tqdm\n",
"from datasets import load_dataset\n",
"from datasets.arrow_dataset import Dataset\n",
- "from transformers import AutoTokenizer, AutoModel, AdamW\n",
- "from transformers.models.bert.modeling_bert import BertPreTrainedModel, BertForSequenceClassification\n",
+ "from transformers import AutoTokenizer, AutoModelForSequenceClassification, AdamW\n",
"from torch.utils.data import DataLoader"
]
},
{
"cell_type": "markdown",
- "id": "smaller-truth",
+ "id": "mineral-justice",
"metadata": {},
"source": [
"# GPU 사용을 위해 device를 설정합니다."
@@ -122,7 +121,7 @@
{
"cell_type": "code",
"execution_count": 3,
- "id": "musical-concord",
+ "id": "pediatric-monday",
"metadata": {},
"outputs": [],
"source": [
@@ -131,7 +130,7 @@
},
{
"cell_type": "markdown",
- "id": "valued-delta",
+ "id": "earned-thirty",
"metadata": {},
"source": [
"# KLUE RE 데이터셋을 가져옵니다."
@@ -140,7 +139,7 @@
{
"cell_type": "code",
"execution_count": 4,
- "id": "induced-breast",
+ "id": "proprietary-pioneer",
"metadata": {},
"outputs": [
{
@@ -157,7 +156,7 @@
},
{
"cell_type": "markdown",
- "id": "higher-compact",
+ "id": "modern-rendering",
"metadata": {},
"source": [
"# 데이터셋은 train과 validation 데이터로 구성되어 있습니다. \n",
@@ -169,7 +168,7 @@
{
"cell_type": "code",
"execution_count": 5,
- "id": "respective-labor",
+ "id": "dried-fellowship",
"metadata": {},
"outputs": [
{
@@ -177,11 +176,11 @@
"text/plain": [
"DatasetDict({\n",
" train: Dataset({\n",
- " features: ['guid', 'label', 'object_entity', 'sentence', 'source', 'subject_entity'],\n",
+ " features: ['guid', 'sentence', 'subject_entity', 'object_entity', 'label', 'source'],\n",
" num_rows: 32470\n",
" })\n",
" validation: Dataset({\n",
- " features: ['guid', 'label', 'object_entity', 'sentence', 'source', 'subject_entity'],\n",
+ " features: ['guid', 'sentence', 'subject_entity', 'object_entity', 'label', 'source'],\n",
" num_rows: 7765\n",
" })\n",
"})"
@@ -198,7 +197,7 @@
},
{
"cell_type": "markdown",
- "id": "increasing-clarity",
+ "id": "geological-engagement",
"metadata": {},
"source": [
"# 데이터 구성\n",
@@ -208,24 +207,24 @@
{
"cell_type": "code",
"execution_count": 6,
- "id": "assisted-brooklyn",
+ "id": "bearing-budapest",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{'guid': 'klue-re-v1_train_00000',\n",
- " 'label': 0,\n",
- " 'object_entity': {'word': '조지 해리슨',\n",
- " 'start_idx': 13,\n",
- " 'end_idx': 18,\n",
- " 'type': 'PER'},\n",
" 'sentence': '〈Something〉는 조지 해리슨이 쓰고 비틀즈가 1969년 앨범 《Abbey Road》에 담은 노래다.',\n",
- " 'source': 'wikipedia',\n",
" 'subject_entity': {'word': '비틀즈',\n",
" 'start_idx': 24,\n",
" 'end_idx': 26,\n",
- " 'type': 'ORG'}}"
+ " 'type': 'ORG'},\n",
+ " 'object_entity': {'word': '조지 해리슨',\n",
+ " 'start_idx': 13,\n",
+ " 'end_idx': 18,\n",
+ " 'type': 'PER'},\n",
+ " 'label': 0,\n",
+ " 'source': 'wikipedia'}"
]
},
"execution_count": 6,
@@ -239,7 +238,7 @@
},
{
"cell_type": "markdown",
- "id": "established-platform",
+ "id": "irish-shoulder",
"metadata": {},
"source": [
"# train데이터에서 각 label의 수를 살펴보겠습니다.\n",
@@ -256,7 +255,7 @@
{
"cell_type": "code",
"execution_count": 7,
- "id": "english-royalty",
+ "id": "miniature-token",
"metadata": {},
"outputs": [
{
@@ -315,7 +314,7 @@
},
{
"cell_type": "markdown",
- "id": "accredited-weapon",
+ "id": "broke-basket",
"metadata": {},
"source": [
"# 개체 양 끝에 special token을 추가합니다.\n",
@@ -333,7 +332,7 @@
{
"cell_type": "code",
"execution_count": 8,
- "id": "surface-squad",
+ "id": "creative-asbestos",
"metadata": {},
"outputs": [],
"source": [
@@ -369,7 +368,7 @@
{
"cell_type": "code",
"execution_count": 9,
- "id": "efficient-absorption",
+ "id": "threaded-client",
"metadata": {},
"outputs": [],
"source": [
@@ -381,7 +380,7 @@
{
"cell_type": "code",
"execution_count": 10,
- "id": "exclusive-travel",
+ "id": "bored-replication",
"metadata": {},
"outputs": [
{
@@ -409,7 +408,7 @@
},
{
"cell_type": "markdown",
- "id": "pointed-earthquake",
+ "id": "auburn-logic",
"metadata": {},
"source": [
"# klue/bert-base 모델을 사용할 예정이므로 모델에 맞는 tokenizer를 가져옵니다."
@@ -418,7 +417,7 @@
{
"cell_type": "code",
"execution_count": 11,
- "id": "southwest-retrieval",
+ "id": "beginning-penetration",
"metadata": {},
"outputs": [],
"source": [
@@ -428,7 +427,7 @@
{
"cell_type": "code",
"execution_count": 12,
- "id": "color-footwear",
+ "id": "initial-drawing",
"metadata": {},
"outputs": [],
"source": [
@@ -437,7 +436,7 @@
},
{
"cell_type": "markdown",
- "id": "recreational-given",
+ "id": "opening-links",
"metadata": {},
"source": [
"# tokenizer를 이용한 토큰화 결과가 어떻게 나오는지 살펴보도록 하겠습니다.\n",
@@ -447,7 +446,7 @@
{
"cell_type": "code",
"execution_count": 13,
- "id": "exact-senegal",
+ "id": "french-nitrogen",
"metadata": {},
"outputs": [],
"source": [
@@ -457,7 +456,7 @@
{
"cell_type": "code",
"execution_count": 14,
- "id": "fatty-finance",
+ "id": "improved-tourist",
"metadata": {},
"outputs": [
{
@@ -478,7 +477,7 @@
{
"cell_type": "code",
"execution_count": 15,
- "id": "electoral-witch",
+ "id": "answering-replica",
"metadata": {},
"outputs": [],
"source": [
@@ -490,7 +489,7 @@
},
{
"cell_type": "markdown",
- "id": "metropolitan-bathroom",
+ "id": "ranging-reception",
"metadata": {},
"source": [
"토큰화 결과로 Bert모델의 입력으로 필요한 input_ids, token_type_ids, attention_mask가 나오는 것을 확인할 수 있습니다.\n",
@@ -501,7 +500,7 @@
{
"cell_type": "code",
"execution_count": 16,
- "id": "pretty-luther",
+ "id": "fixed-heater",
"metadata": {},
"outputs": [
{
@@ -521,7 +520,7 @@
},
{
"cell_type": "markdown",
- "id": "wrong-corpus",
+ "id": "patent-korea",
"metadata": {},
"source": [
"토큰화 된 문장을 다시 디코딩 해봄으로써 원본 문장을 얻을 수 있는지 확인해봅니다.\n",
@@ -532,7 +531,7 @@
{
"cell_type": "code",
"execution_count": 17,
- "id": "cellular-arrow",
+ "id": "becoming-browser",
"metadata": {},
"outputs": [
{
@@ -552,7 +551,7 @@
},
{
"cell_type": "markdown",
- "id": "variable-channels",
+ "id": "sunset-questionnaire",
"metadata": {},
"source": [
"# Special token 추가\n",
@@ -564,7 +563,7 @@
{
"cell_type": "code",
"execution_count": 18,
- "id": "matched-investor",
+ "id": "positive-blame",
"metadata": {},
"outputs": [],
"source": [
@@ -574,7 +573,7 @@
},
{
"cell_type": "markdown",
- "id": "secondary-paris",
+ "id": "unlike-terminology",
"metadata": {},
"source": [
"데이터로더 및 학습에 필요한 값들을 설정합니다."
@@ -583,7 +582,7 @@
{
"cell_type": "code",
"execution_count": 19,
- "id": "developmental-intersection",
+ "id": "assumed-parts",
"metadata": {},
"outputs": [],
"source": [
@@ -601,7 +600,7 @@
},
{
"cell_type": "markdown",
- "id": "literary-engagement",
+ "id": "nearby-packet",
"metadata": {},
"source": [
"# 학습에 이용할 데이터셋과 데이터로더를 만들어 줍니다."
@@ -610,7 +609,7 @@
{
"cell_type": "code",
"execution_count": 20,
- "id": "trying-consent",
+ "id": "domestic-hampshire",
"metadata": {},
"outputs": [],
"source": [
@@ -635,7 +634,7 @@
{
"cell_type": "code",
"execution_count": 21,
- "id": "thick-effectiveness",
+ "id": "tracked-regression",
"metadata": {},
"outputs": [],
"source": [
@@ -648,7 +647,7 @@
},
{
"cell_type": "markdown",
- "id": "electronic-payday",
+ "id": "virtual-vertex",
"metadata": {},
"source": [
"# klue/bert-base 모델을 로드합니다.\n",
@@ -669,7 +668,7 @@
{
"cell_type": "code",
"execution_count": 22,
- "id": "rocky-justice",
+ "id": "refined-cornell",
"metadata": {},
"outputs": [
{
@@ -685,12 +684,12 @@
}
],
"source": [
- "model = BertForSequenceClassification.from_pretrained(model_name, num_labels=num_labels).to(device)"
+ "model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=num_labels).to(device)"
]
},
{
"cell_type": "markdown",
- "id": "electronic-asian",
+ "id": "enormous-germany",
"metadata": {},
"source": [
"모델 구성의 마지막에 있는 classifier 부분을 보시면 출력 차원이 30으로 설정되어 있는 것을 확인할 수 있습니다."
@@ -699,7 +698,7 @@
{
"cell_type": "code",
"execution_count": 23,
- "id": "attached-particle",
+ "id": "congressional-south",
"metadata": {},
"outputs": [
{
@@ -1015,7 +1014,7 @@
},
{
"cell_type": "markdown",
- "id": "alternate-import",
+ "id": "modular-party",
"metadata": {},
"source": [
"# Bert Embedding Layer을 resize합니다.\n",
@@ -1032,7 +1031,7 @@
{
"cell_type": "code",
"execution_count": 24,
- "id": "stunning-worthy",
+ "id": "editorial-johnston",
"metadata": {},
"outputs": [
{
@@ -1052,7 +1051,7 @@
},
{
"cell_type": "markdown",
- "id": "promising-yesterday",
+ "id": "polar-intention",
"metadata": {},
"source": [
"학습 도중 Loss, Accuracy 계산 및 저장을 간단하게 하기 위해 AverageMeter를 클래스를 이용합니다."
@@ -1061,7 +1060,7 @@
{
"cell_type": "code",
"execution_count": 25,
- "id": "hourly-bridges",
+ "id": "numerical-pitch",
"metadata": {},
"outputs": [],
"source": [
@@ -1081,7 +1080,7 @@
},
{
"cell_type": "markdown",
- "id": "laughing-electricity",
+ "id": "developed-qatar",
"metadata": {},
"source": [
"# Model fine-tuning\n",
@@ -1095,7 +1094,7 @@
{
"cell_type": "code",
"execution_count": 26,
- "id": "silver-algeria",
+ "id": "modern-ancient",
"metadata": {},
"outputs": [
{
@@ -1116,7 +1115,7 @@
"name": "stderr",
"output_type": "stream",
"text": [
- "100%|██████████| 4059/4059 [14:11<00:00, 4.77it/s]\n",
+ "100%|██████████| 4059/4059 [14:15<00:00, 4.74it/s]\n",
"100%|██████████| 971/971 [01:07<00:00, 14.30it/s]\n",
" 0%| | 0/4059 [00:00\n"
]
@@ -1134,8 +1133,8 @@
"name": "stderr",
"output_type": "stream",
"text": [
- "100%|██████████| 4059/4059 [14:10<00:00, 4.77it/s]\n",
- "100%|██████████| 971/971 [01:07<00:00, 14.30it/s]\n",
+ "100%|██████████| 4059/4059 [14:17<00:00, 4.73it/s]\n",
+ "100%|██████████| 971/971 [01:08<00:00, 14.18it/s]\n",
" 0%| | 0/4059 [00:00\n"
]
@@ -1152,15 +1151,15 @@
"name": "stderr",
"output_type": "stream",
"text": [
- "100%|██████████| 4059/4059 [14:10<00:00, 4.77it/s]\n",
- "100%|██████████| 971/971 [01:07<00:00, 14.32it/s]"
+ "100%|██████████| 4059/4059 [14:16<00:00, 4.74it/s]\n",
+ "100%|██████████| 971/971 [01:08<00:00, 14.17it/s]"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
- "train_loss: 0.3608, train_acc: 87.74%, val_loss: 0.7518, val_acc: 75.54%\n",
+ "train_loss: 0.3583, train_acc: 0.8792, val_loss: 0.8018, val_acc: 0.7415\n",
"====================================================================================================\n"
]
},
@@ -1204,7 +1203,7 @@
" \n",
" results = {\n",
" 'loss': loss_save.avg,\n",
- " 'acc': acc_save.avg * 100,\n",
+ " 'acc': acc_save.avg,\n",
" }\n",
" \n",
" return results\n",
@@ -1231,13 +1230,13 @@
" val_loss, val_acc = val_results['loss'], val_results['acc']\n",
" \n",
" \n",
- " print(f'train_loss: {train_loss:.4f}, train_acc: {train_acc:.2f}%, val_loss: {val_loss:.4f}, val_acc: {val_acc:.2f}%')\n",
+ " print(f'train_loss: {train_loss:.4f}, train_acc: {train_acc:.4f}, val_loss: {val_loss:.4f}, val_acc: {val_acc:.4f}')\n",
" print('=' * 100)"
]
},
{
"cell_type": "markdown",
- "id": "ongoing-tooth",
+ "id": "steady-climb",
"metadata": {},
"source": [
"# 학습된 모델을 저장합니다."
@@ -1246,7 +1245,7 @@
{
"cell_type": "code",
"execution_count": 27,
- "id": "consistent-cassette",
+ "id": "built-petite",
"metadata": {},
"outputs": [],
"source": [
@@ -1256,7 +1255,7 @@
},
{
"cell_type": "markdown",
- "id": "nervous-positive",
+ "id": "amended-debut",
"metadata": {},
"source": [
"# Validation 결과 확인\n",
@@ -1266,7 +1265,7 @@
{
"cell_type": "code",
"execution_count": 28,
- "id": "small-planet",
+ "id": "expanded-shopping",
"metadata": {},
"outputs": [
{
@@ -1289,7 +1288,7 @@
{
"cell_type": "code",
"execution_count": 29,
- "id": "pediatric-wednesday",
+ "id": "retained-works",
"metadata": {},
"outputs": [],
"source": [
@@ -1303,7 +1302,7 @@
{
"cell_type": "code",
"execution_count": 30,
- "id": "assigned-slovak",
+ "id": "fleet-kidney",
"metadata": {},
"outputs": [],
"source": [
@@ -1320,7 +1319,7 @@
},
{
"cell_type": "markdown",
- "id": "prime-device",
+ "id": "configured-chase",
"metadata": {},
"source": [
"0번 라벨은 \"관계 없음\"이므로 해당 문장에 대해서 예측이 정확하게 됐음을 알 수 있습니다."
@@ -1329,7 +1328,7 @@
{
"cell_type": "code",
"execution_count": 31,
- "id": "indoor-stuff",
+ "id": "million-mailing",
"metadata": {},
"outputs": [
{
@@ -1349,7 +1348,7 @@
},
{
"cell_type": "markdown",
- "id": "leading-rebecca",
+ "id": "educated-receiver",
"metadata": {},
"source": [
"다음으로 관계가 있는 문장에 대한 예측 결과를 살펴보겠습니다."
@@ -1358,7 +1357,7 @@
{
"cell_type": "code",
"execution_count": 32,
- "id": "architectural-engagement",
+ "id": "little-showcase",
"metadata": {},
"outputs": [
{
@@ -1381,7 +1380,7 @@
{
"cell_type": "code",
"execution_count": 33,
- "id": "enormous-shift",
+ "id": "local-cancer",
"metadata": {},
"outputs": [],
"source": [
@@ -1395,7 +1394,7 @@
{
"cell_type": "code",
"execution_count": 34,
- "id": "aware-savage",
+ "id": "interpreted-airport",
"metadata": {},
"outputs": [],
"source": [
@@ -1412,7 +1411,7 @@
},
{
"cell_type": "markdown",
- "id": "found-option",
+ "id": "particular-compensation",
"metadata": {},
"source": [
"3번 라벨은 \"org:place_of_headquarters\" 이므로 해당 문장에 대해서 예측이 정확하게 됐음을 알 수 있습니다."
@@ -1421,7 +1420,7 @@
{
"cell_type": "code",
"execution_count": 35,
- "id": "advanced-creek",
+ "id": "domestic-vietnam",
"metadata": {},
"outputs": [
{
@@ -1441,7 +1440,7 @@
},
{
"cell_type": "markdown",
- "id": "continuous-professor",
+ "id": "wrapped-invite",
"metadata": {},
"source": [
"# 모델 평가\n",
@@ -1455,7 +1454,7 @@
{
"cell_type": "code",
"execution_count": 36,
- "id": "surgical-hours",
+ "id": "wrapped-phoenix",
"metadata": {},
"outputs": [],
"source": [
@@ -1479,14 +1478,14 @@
{
"cell_type": "code",
"execution_count": 37,
- "id": "literary-granny",
+ "id": "metric-cruise",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
- "100%|██████████| 971/971 [01:07<00:00, 14.45it/s]\n"
+ "100%|██████████| 971/971 [01:07<00:00, 14.32it/s]\n"
]
}
],
@@ -1517,22 +1516,22 @@
},
{
"cell_type": "markdown",
- "id": "functional-cotton",
+ "id": "theoretical-background",
"metadata": {},
"source": [
- "약 63.58의 F1 score를 기록했습니다. 하이퍼파라미터, 모델 구조 등을 변경시켜 더 높은 F1 score를 기록해보세요!"
+ "약 61.93의 F1 score를 기록했습니다. 하이퍼파라미터, 모델 구조 등을 변경시켜 다양한 학습을 시도해보세요!"
]
},
{
"cell_type": "code",
"execution_count": 38,
- "id": "general-attitude",
+ "id": "imposed-globe",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
- "63.58118361153262"
+ "61.93164001254312"
]
},
"execution_count": 38,
@@ -1546,7 +1545,7 @@
},
{
"cell_type": "markdown",
- "id": "bigger-action",
+ "id": "local-insulation",
"metadata": {},
"source": [
"학습된 모델은 Huggingface Model Hub에 배포되어 있고, 언제든지 다운받아서 활용이 가능합니다.\n",
@@ -1556,7 +1555,7 @@
},
{
"cell_type": "markdown",
- "id": "capable-found",
+ "id": "shared-dating",
"metadata": {},
"source": [
"# Reference"
@@ -1564,7 +1563,7 @@
},
{
"cell_type": "markdown",
- "id": "noble-first",
+ "id": "drawn-trick",
"metadata": {},
"source": [
"### klue-transformers-tutorial\n",