MultiClass train 예제¶

종속 변수(y)에 따른 분류¶

Binary Class
- 단순히 0 또는 1 [0,1,0,...]
Multi Class
- 0 ~ k개까지 존재하고 이 중 한가지 class로 분류하는 것. [[0,1,0],[1,0,0],...]
Multi Label
- 0 ~ k까지 존재하며, 이 중 최대 k개 까지 중복된 class가 될 수 있는 것. [[1,1,0],[1,0,1],...]

multiclass를 위해서 적절한 데이터를 찾기가 힘들었습니다. 샘플이니 데이터가 좀 작았으면 해서 Kopora의 korean_hate_speech data를 사용하였습니다.

import torch
from Korpora import Korpora
import pandas as pd

fineturning할 Multiclass 데이터를 가져옵니다. hate 항목이 3가지로 나뉨, 적절한 예제로 선택함

korean_hate_speech = Korpora.load('korean_hate_speech')

    Korpora 는 다른 분들이 연구 목적으로 공유해주신 말뭉치들을
    손쉽게 다운로드, 사용할 수 있는 기능만을 제공합니다.

    말뭉치들을 공유해 주신 분들에게 감사드리며, 각 말뭉치 별 설명과 라이센스를 공유 드립니다.
    해당 말뭉치에 대해 자세히 알고 싶으신 분은 아래의 description 을 참고,
    해당 말뭉치를 연구/상용의 목적으로 이용하실 때에는 아래의 라이센스를 참고해 주시기 바랍니다.

    # Description
    Authors :
        - Jihyung Moon* (inmoonlight@github)
        - Won Ik Cho* (warnikchow@github)
        - Junbum Lee (beomi@github)
        * equal contribution
    Repository : https://github.com/kocohub/korean-hate-speech
    References :
        - Moon, J., Cho, W. I., & Lee, J. (2020). BEEP! Korean Corpus of Online News
          Comments for Toxic Speech Detection. arXiv preprint arXiv:2005.12503.

    We provide the first human-annotated Korean corpus for toxic speech detection and the large unlabeled corpus.
    The data is comments from the Korean entertainment news aggregation platform.

    # License
    Creative Commons Attribution-ShareAlike 4.0 International License.
    Visit here for detail : https://creativecommons.org/licenses/by-sa/4.0/

[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\unlabeled\unlabeled_comments_1.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\unlabeled\unlabeled_comments_2.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\unlabeled\unlabeled_comments_3.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\unlabeled\unlabeled_comments_4.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\unlabeled\unlabeled_comments_5.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\unlabeled_comments.news_title_1.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\unlabeled_comments.news_title_2.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\unlabeled_comments.news_title_3.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\unlabeled_comments.news_title_4.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\unlabeled_comments.news_title_5.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\dev.news_title.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\test.news_title.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\news_title\train.news_title.txt
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\labeled\dev.tsv
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\labeled\train.tsv
[Korpora] Corpus `korean hate speech` is already installed at C:\Users\jun\Korpora\korean_hate_speech\test.no_label.tsv

korean_hate_speech.train

KoreanHateSpeech.train: size=7896
  - KoreanHateSpeech.train.texts : list[str]
  - KoreanHateSpeech.train.titles : list[str]
  - KoreanHateSpeech.train.gender_biases : list[str]
  - KoreanHateSpeech.train.biases : list[str]
  - KoreanHateSpeech.train.hates : list[str]

korean_hate_speech.dev

KoreanHateSpeech.dev: size=471
  - KoreanHateSpeech.dev.texts : list[str]
  - KoreanHateSpeech.dev.titles : list[str]
  - KoreanHateSpeech.dev.gender_biases : list[str]
  - KoreanHateSpeech.dev.biases : list[str]
  - KoreanHateSpeech.dev.hates : list[str]

dataframe 에 넣어봅니다.

train_data = pd.DataFrame({"texts":korean_hate_speech.train.texts, "titles":korean_hate_speech.train.titles,
                          "gender_biases":korean_hate_speech.train.gender_biases, "biases":korean_hate_speech.train.biases,
                          "hates":korean_hate_speech.train.hates})
test_data = pd.DataFrame({"texts":korean_hate_speech.dev.texts, "titles":korean_hate_speech.dev.titles,
                          "gender_biases":korean_hate_speech.dev.gender_biases, "biases":korean_hate_speech.dev.biases,
                          "hates":korean_hate_speech.dev.hates})

train_data

	texts	titles	gender_biases	biases	hates
0	(현재 호텔주인 심정) 아18 난 마른하늘에 날벼락맞고 호텔망하게생겼는데 누군 계속...	"밤새 조문 행렬…故 전미선, 동료들이 그리워하는 따뜻한 배우 [종합]"	False	others	hate
1	....한국적인 미인의 대표적인 분...너무나 곱고아름다운모습...그모습뒤의 슬픔을...	"'연중' 故 전미선, 생전 마지막 미공개 인터뷰…환하게 웃는 모습 '먹먹'[종합]"	False	none	none
2	...못된 넘들...남의 고통을 즐겼던 넘들..이젠 마땅한 처벌을 받아야지..,그래...	"[단독] 잔나비, 라디오 출연 취소→'한밤' 방송 연기..비판 여론 ing(종합)"	False	none	hate
3	1,2화 어설펐는데 3,4화 지나서부터는 갈수록 너무 재밌던데	"'아스달 연대기' 장동건-김옥빈, 들끓는 '욕망커플'→눈물범벅 '칼끝 대립'"	False	none	none
4	1. 사람 얼굴 손톱으로 긁은것은 인격살해이고2. 동영상이 몰카냐? 메걸리안들 생각...	[DA:이슈] ‘구하라 비보’ 최종범 항소심에 영향?…법조계 “‘공소권 없음’ 아냐”	True	gender	hate
...	...	...	...	...	...
7891	힘내세요~ 응원합니다!!	"허지웅, 허투루 넘길 말 없었다…솔직하게 드러냈던 속사정"	False	none	none
7892	힘내세요~~삼가 고인의 명복을 빕니다..	"이혜경, ‘오! 캐롤’ 공연 중 남편 오정욱 부고…오열 속 발인 [종합]"	False	none	none
7893	힘내세용 ^^ 항상 응원합니닷 ^^ !	"'설경구♥' 송윤아, 아들과 즐거운 하루 ""전 엄마니까요"""	False	none	none
7894	힘내소...연기로 답해요.나도 53살 인데 이런일 저런일 다 있더라구요.인격을 믿습...	"[SC현장]""연예인 인생 협박 유감""…미소잃은 최민수, '보복운전 혐의' 2차...	False	none	none
7895	힘들면 관뒀어야지 그게 현명한거다	"[단독]스태프 사망사고 '서른이지만', 결국 오늘 촬영 취소"	False	none	none

7896 rows × 5 columns

test_data

	texts	titles	gender_biases	biases	hates
0	송중기 시대극은 믿고본다. 첫회 신선하고 좋았다.	"'아스달 연대기' 뇌안탈 후계자 송중기, 이아르크에 숨어들었다 [종합]"	False	none	none
1	지현우 나쁜놈	‘섹션TV’ 유인나 “결혼은 3년 후쯤 했으면…노력해보겠다”	False	none	offensive
2	알바쓰고많이만들면되지 돈욕심없으면골목식당왜나온겨 기댕기게나하고 산에가서팔어라	"'골목식당' 포방터시장, 계속되는 논란의 중심[SS초점]"	False	none	hate
3	설마 ㅈ 현정 작가 아니지??	"""마음 고생""…나영석·정유미, 황당 불륜 지라시→방송작가 검거→벌금형 선고 [...	True	gender	hate
4	이미자씨 송혜교씨 돈이 그리 많으면 탈세말고 그돈으로 평소에 불우이웃에게 기부도 좀...	"이미자 탈세, 10년간 44억 넘는 소득 신고 누락…약 20억 세금 추가 납부"	False	none	offensive
...	...	...	...	...	...
466	지현우 범죄 저지르지 않았나요?	"[SC현장] '사생결단' 이시영♥지현우, 8년만의 호르몬 로맨스 통할까(종합)"	False	none	offensive
467	여자인생 망칠 일 있나 ㅋㅋ	"박성광, 5월 2일 결혼...""예비신부는 7살 연하 비연예인""(공식)"	True	gender	hate
468	근데 전라도에서 사고가 났는데 굳이 서울까지 와서 병원에 가느 이유는?	"[POP이슈]""차량 반파 교통사고""…송가인, 목·허리 통증→정밀검사 진행(종합)"	False	others	offensive
469	할매젖x, 뱃살x, 몸매 s라인, 유륜은 적당해야됨(너무크거나 너무 작아도 x), ...	"[인터뷰①] 수애 ""노출·베드신多 장르, 부담보다 도전이라 생각했다"""	True	gender	hate
470	남자가 잘못한거라면... 반성도 없다면...나였다면 ... 여자처럼 아주 못되게 할...	"'안재현과 갈등' 구혜선, SNS 활동 재개…""다시 시작"" [종합]"	True	gender	none

471 rows × 5 columns

train_data[['hates']].value_counts()

hates    
none         3486
offensive    2499
hate         1911
dtype: int64

max(len(l) for l in train_data['texts'])

max(len(l) for l in test_data['texts'])

학습에 사용될 pre-trained 된 BERT 모델을 가져와서 토큰화 하기

pretrained_model_name="beomi/kcbert-base"

from transformers import AutoTokenizer

# 경고가 뜬다면 다음 명령으로 설치해주자 !pip install ipywidgets

tokenizer = AutoTokenizer.from_pretrained(
    pretrained_model_name
)

tokenized_train_sentences = tokenizer(
    list(train_data.texts),
    return_tensors="pt",
    padding=True,
    truncation=True,
)

tokenized_test_sentences = tokenizer(
    list(test_data.texts),
    return_tensors="pt",
    padding=True,
    truncation=True,
)

출력해봅니다.

print(tokenized_train_sentences.keys())
print(tokenized_train_sentences['input_ids'])
print(tokenized_train_sentences['attention_mask'])
print(tokenized_train_sentences['token_type_ids'])

dict_keys(['input_ids', 'token_type_ids', 'attention_mask'])
tensor([[    2,    11,  8979,  ...,     0,     0,     0],
        [    2,    17,    17,  ...,     0,     0,     0],
        [    2,    17,    17,  ...,     0,     0,     0],
        ...,
        [    2,  9104,  4066,  ...,     0,     0,     0],
        [    2,  9104,  4266,  ...,     0,     0,     0],
        [    2, 24825,   323,  ...,     0,     0,     0]])
tensor([[1, 1, 1,  ..., 0, 0, 0],
        [1, 1, 1,  ..., 0, 0, 0],
        [1, 1, 1,  ..., 0, 0, 0],
        ...,
        [1, 1, 1,  ..., 0, 0, 0],
        [1, 1, 1,  ..., 0, 0, 0],
        [1, 1, 1,  ..., 0, 0, 0]])
tensor([[0, 0, 0,  ..., 0, 0, 0],
        [0, 0, 0,  ..., 0, 0, 0],
        [0, 0, 0,  ..., 0, 0, 0],
        ...,
        [0, 0, 0,  ..., 0, 0, 0],
        [0, 0, 0,  ..., 0, 0, 0],
        [0, 0, 0,  ..., 0, 0, 0]])

tokenized_train_sentences['input_ids'].shape

torch.Size([7896, 74])

train_labels=pd.get_dummies(train_data[['hates']])
train_labels # hate:[1,0,0], none:[0,1,0], offensive:[0,0,1]

	hates_hate	hates_none	hates_offensive
0	1	0	0
1	0	1	0
2	1	0	0
3	0	1	0
4	1	0	0
...	...	...	...
7891	0	1	0
7892	0	1	0
7893	0	1	0
7894	0	1	0
7895	0	1	0

7896 rows × 3 columns

one hot encoding 해줍니다.

test_labels=pd.get_dummies(test_data[['hates']])
test_labels

	hates_hate	hates_none	hates_offensive
0	0	1	0
1	0	0	1
2	1	0	0
3	1	0	0
4	0	0	1
...	...	...	...
466	0	0	1
467	1	0	0
468	0	0	1
469	1	0	0
470	0	1	0

471 rows × 3 columns

train_labels.values

array([[1, 0, 0],
       [0, 1, 0],
       [1, 0, 0],
       ...,
       [0, 1, 0],
       [0, 1, 0],
       [0, 1, 0]])

test_labels.values

array([[0, 1, 0],
       [0, 0, 1],
       [1, 0, 0],
       ...,
       [0, 0, 1],
       [1, 0, 0],
       [0, 1, 0]])

멀티 Class 일때 label shape가 제대로 생성되었는지 확인하도록 한다.

train_label = train_labels.values.astype(float) # 꼭 float 로 변환해 줍니다.
test_label = test_labels.values.astype(float)

데이터 로더 준비, 이것이 필요한 이유는 배치 처리하는 내부에서 원소를 액세스 하기 위함

class DataloaderDataset(torch.utils.data.Dataset):
    def __init__(self, encodings, labels):
        self.encodings = encodings
        self.labels = labels

    def __getitem__(self, idx):
        item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
        item['labels'] = torch.tensor(self.labels[idx])
        return item

    def __len__(self):
        return len(self.labels)

train_dataset = DataloaderDataset(tokenized_train_sentences, train_label)
test_dataset = DataloaderDataset(tokenized_test_sentences, test_label)

from transformers import BertConfig, AutoModelForSequenceClassification, Trainer, TrainingArguments

device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
print(device)

cuda:0

pretrained_model_config = BertConfig.from_pretrained(
    pretrained_model_name,
)
model = AutoModelForSequenceClassification.from_pretrained(
        pretrained_model_name,
        #config=pretrained_model_config,
        num_labels=3,
        #problem_type="multi_label_classification",
)

Some weights of the model checkpoint at beomi/kcbert-base were not used when initializing BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.bias', 'cls.predictions.decoder.bias', 'cls.predictions.transform.dense.weight', 'cls.seq_relationship.weight']
- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of BertForSequenceClassification were not initialized from the model checkpoint at beomi/kcbert-base and are newly initialized: ['classifier.bias', 'classifier.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.

labels 에 float 형이 들어가 있으므로 np.argmax 로 동일하게 정수형으로 출력되도록 해줍니다. labels_ = np.argmax(labels, axis=-1)

#!pip install evaluate
#!pip install scikit-learn
import numpy as np
import evaluate 
metric = evaluate.load("accuracy")
def compute_metrics(eval_pred):
    logits, labels = eval_pred
    predictions = np.argmax(logits, axis=-1)
    labels_ = np.argmax(labels, axis=-1)
    return metric.compute(predictions=predictions, references=labels_)

training_args = TrainingArguments(
    output_dir='./results',          # output directory
    num_train_epochs=1,              # total number of training epochs
    #per_device_train_batch_size=32,  # batch size per device during training
    #per_device_eval_batch_size=64,   # batch size for evaluation
    per_device_train_batch_size=5,  # batch size per device during training
    per_device_eval_batch_size=5,   # batch size for evaluation
    warmup_steps=500,                # number of warmup steps for learning rate scheduler
    weight_decay=0.01,               # strength of weight decay
    logging_dir='./logs',            # directory for storing logs
    logging_steps=100,
    save_steps=200,
    save_total_limit=2,
    save_on_each_node=True,
    do_train=True,                   # Perform training
    do_eval=True,                    # Perform evaluation
    evaluation_strategy="epoch",
    seed=3
)

https://huggingface.co/docs/transformers/v4.19.2/en/main_classes/trainer#transformers.TrainingArguments

trainer = Trainer(
    model=model,                         # the instantiated 🤗 Transformers model to be trained
    args=training_args,                  # training arguments, defined above
    train_dataset=train_dataset,         # training dataset
    eval_dataset=test_dataset,
    compute_metrics=compute_metrics,
)

trainer.train()

***** Running training *****
  Num examples = 7896
  Num Epochs = 1
  Instantaneous batch size per device = 5
  Total train batch size (w. parallel, distributed & accumulation) = 5
  Gradient Accumulation steps = 1
  Total optimization steps = 1580
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}

[1580/1580 02:41, Epoch 1/1]

Epoch	Training Loss	Validation Loss	Accuracy
1	0.489900	0.463366	0.683652

Saving model checkpoint to ./results\checkpoint-200
Configuration saved in ./results\checkpoint-200\config.json
Model weights saved in ./results\checkpoint-200\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-1200] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
Saving model checkpoint to ./results\checkpoint-400
Configuration saved in ./results\checkpoint-400\config.json
Model weights saved in ./results\checkpoint-400\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-1400] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
Saving model checkpoint to ./results\checkpoint-600
Configuration saved in ./results\checkpoint-600\config.json
Model weights saved in ./results\checkpoint-600\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-200] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
Saving model checkpoint to ./results\checkpoint-800
Configuration saved in ./results\checkpoint-800\config.json
Model weights saved in ./results\checkpoint-800\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-400] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
Saving model checkpoint to ./results\checkpoint-1000
Configuration saved in ./results\checkpoint-1000\config.json
Model weights saved in ./results\checkpoint-1000\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-600] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
Saving model checkpoint to ./results\checkpoint-1200
Configuration saved in ./results\checkpoint-1200\config.json
Model weights saved in ./results\checkpoint-1200\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-800] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
Saving model checkpoint to ./results\checkpoint-1400
Configuration saved in ./results\checkpoint-1400\config.json
Model weights saved in ./results\checkpoint-1400\pytorch_model.bin
Deleting older checkpoint [results\checkpoint-1000] due to args.save_total_limit
C:\Users\jun\AppData\Local\Temp\ipykernel_29796\1263192275.py:7: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
***** Running Evaluation *****
  Num examples = 471
  Batch size = 5


Training completed. Do not forget to share your model on huggingface.co/models =)

TrainOutput(global_step=1580, training_loss=0.5128312267834627, metrics={'train_runtime': 163.0519, 'train_samples_per_second': 48.426, 'train_steps_per_second': 9.69, 'total_flos': 300269965687776.0, 'train_loss': 0.5128312267834627, 'epoch': 1.0})

1 epoch에 정확도가 68%정도 나옵니다.

trainer.save_model("trained_model_hate")

Saving model checkpoint to trained_model_hate
Configuration saved in trained_model_hate\config.json
Model weights saved in trained_model_hate\pytorch_model.bin

환경

!pip freeze

absl-py==1.4.0
aiohttp==3.8.4
aiosignal==1.3.1
anyio==3.6.2
argon2-cffi==21.3.0
argon2-cffi-bindings==21.2.0
arrow==1.2.3
asttokens==2.2.1
async-timeout==4.0.2
attrs==22.2.0
backcall==0.2.0
beautifulsoup4==4.11.2
bleach==6.0.0
cachetools==5.3.0
certifi==2022.12.7
cffi==1.15.1
charset-normalizer==3.1.0
click==8.1.3
colorama==0.4.6
comm==0.1.2
dataclasses==0.6
datasets==2.10.1
debugpy==1.6.6
decorator==5.1.1
defusedxml==0.7.1
dill==0.3.6
evaluate==0.4.0
executing==1.2.0
fastjsonschema==2.16.3
filelock==3.9.0
Flask==2.2.3
Flask-Cors==3.0.10
flask-ngrok==0.0.25
fqdn==1.5.1
frozenlist==1.3.3
fsspec==2023.3.0
google-auth==2.16.2
google-auth-oauthlib==0.4.6
grpcio==1.51.3
huggingface-hub==0.13.0
idna==3.4
ipykernel==6.21.3
ipython==8.11.0
ipython-genutils==0.2.0
ipywidgets==8.0.4
isoduration==20.11.0
itsdangerous==2.1.2
jedi==0.18.2
Jinja2==3.1.2
joblib==1.2.0
jsonpointer==2.3
jsonschema==4.17.3
jupyter-events==0.6.3
jupyter_client==8.0.3
jupyter_core==5.2.0
jupyter_server==2.4.0
jupyter_server_terminals==0.4.4
jupyterlab-pygments==0.2.2
jupyterlab-widgets==3.0.5
Korpora==0.2.0
Markdown==3.4.1
MarkupSafe==2.1.2
matplotlib-inline==0.1.6
mistune==2.0.5
multidict==6.0.4
multiprocess==0.70.14
nbclassic==0.5.3
nbclient==0.7.2
nbconvert==7.2.9
nbformat==5.7.3
nest-asyncio==1.5.6
notebook==6.5.3
notebook_shim==0.2.2
numpy==1.24.2
oauthlib==3.2.2
packaging==23.0
pandas==1.5.3
pandocfilters==1.5.0
parso==0.8.3
pickleshare==0.7.5
Pillow==9.4.0
platformdirs==3.1.1
prometheus-client==0.16.0
prompt-toolkit==3.0.38
protobuf==4.22.1
psutil==5.9.4
pure-eval==0.2.2
pyarrow==11.0.0
pyasn1==0.4.8
pyasn1-modules==0.2.8
pycparser==2.21
pyDeprecate==0.3.2
Pygments==2.14.0
pyrsistent==0.19.3
python-dateutil==2.8.2
python-json-logger==2.0.7
pytorch-lightning==1.6.1
pytz==2022.7.1
pywin32==305
pywinpty==2.0.10
PyYAML==6.0
pyzmq==25.0.0
ratsnlp==1.0.52
regex==2022.10.31
requests==2.28.2
requests-oauthlib==1.3.1
responses==0.18.0
rfc3339-validator==0.1.4
rfc3986-validator==0.1.1
rsa==4.9
sacremoses==0.0.53
scikit-learn==1.2.2
scipy==1.10.1
Send2Trash==1.8.0
six==1.16.0
sniffio==1.3.0
soupsieve==2.4
stack-data==0.6.2
tensorboard==2.12.0
tensorboard-data-server==0.7.0
tensorboard-plugin-wit==1.8.1
terminado==0.17.1
threadpoolctl==3.1.0
tinycss2==1.2.1
tokenizers==0.10.3
torch==1.13.1+cu116
torchaudio==0.13.1
torchmetrics==0.11.3
torchvision==0.14.1
tornado==6.2
tqdm==4.65.0
traitlets==5.9.0
transformers==4.10.0
typing_extensions==4.5.0
uri-template==1.2.0
urllib3==1.26.14
wcwidth==0.2.6
webcolors==1.12
webencodings==0.5.1
websocket-client==1.5.1
Werkzeug==2.2.3
widgetsnbextension==4.0.5
xlrd==2.0.1
xxhash==3.2.0
yarl==1.8.2

import platform
platform.python_version()

'3.10.10'

SW정리

2023년 3월 19일 일요일

BERT MultiClass train 예제

MultiClass train 예제¶

종속 변수(y)에 따른 분류¶

댓글 없음:

댓글 쓰기

	hates_hate	hates_none	hates_offensive
0	0	1	0
1	0	0	1
2	1	0	0
3	1	0	0
4	0	0	1
...	...	...	...
466	0	0	1
467	1	0	0
468	0	0	1
469	1	0	0
470	0	1	0

	hates_hate	hates_none	hates_offensive
0	0	1	0
1	0	0	1
2	1	0	0
3	1	0	0
4	0	0	1
...	...	...	...
466	0	0	1
467	1	0	0
468	0	0	1
469	1	0	0
470	0	1	0

	hates_hate	hates_none	hates_offensive
0	0	1	0
1	0	0	1
2	1	0	0
3	1	0	0
4	0	0	1
...	...	...	...
466	0	0	1
467	1	0	0
468	0	0	1
469	1	0	0
470	0	1	0