{
  "learning_rate": 9.78310992630157e-06,
  "num_train_epochs": 8,
  "weight_decay": 0.06436845335086991,
  "warmup_ratio": 0.10859899755289561,
  "per_device_train_batch_size": 32
}

사용 방법

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# 모델 로드
model_name = "FISA-conclave/klue-roberta-news-sentiment"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# 예측
text = "삼성전자의 3분기 실적이 시장 예상을 크게 상회했다."
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)

with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.softmax(outputs.logits, dim=1)[0]
    pred = torch.argmax(probs).item()

labels = {0: "negative", 1: "neutral", 2: "positive"}
print(f"예측: {labels[pred]} ({probs[pred]:.2%})")

학습 데이터

총 샘플 수: 9,992개
출처:
- finance_sentiment_corpus
- korfin-asc
- twice_kr_fin

베이스 모델

klue/roberta-base

인용

bibtex

@misc{klue-roberta-news-sentiment,
  author = {Tobykim},
  title = {KLUE-RoBERTa News Sentiment Analysis},
  year = {2024},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/FISA-conclave/klue-roberta-news-sentiment}}
}

라이센스

Apache 2.0