README.md

---
language:
  - ko
tags:
- generated_from_keras_callback
model-index:
- name: t5-base-korean-news-title-klue-ynat
  results: []
---

# t5-large-korean-text-summary


이 모델은 lcw99 / t5-base-korean-text-summary을 klue-ynat으로 훈련시켜 만든 모델입니다.
Input = ['IT과학','경제','사회','생활문화','세계','스포츠','정치']
OUTPUT = 각 label에 맞는 뉴스 기사 제목을 생성합니다.

## Usage
```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model_dir = "kfkas/t5-base-korean-news-title-klue-ynat"
tokenizer = AutoTokenizer.from_pretrained(model_dir)
model = AutoModelForSeq2SeqLM.from_pretrained(model_dir)
model.to(device)

label_list = ['IT과학','경제','사회','생활문화','세계','스포츠','정치']
text = "IT과학"

inputs = tokenizer.encode(text, max_length=256, truncation=True, return_tensors="pt")
with torch.no_grad():
  output = model.generate(
    input_ids,
    do_sample=True, #샘플링 전략 사용
    max_length=128, # 최대 디코딩 길이는 50
    top_k=50, # 확률 순위가 50위 밖인 토큰은 샘플링에서 제외
    top_p=0.95, # 누적 확률이 95%인 후보집합에서만 생성
)
decoded_output = tokenizer.decode(output, skip_special_tokens=True)[0]
print(predicted_title)
```


## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- optimizer: None
- training_precision: float16
### Training results
### Framework versions
- Transformers 4.22.1
- TensorFlow 2.10.0
- Datasets 2.5.1
- Tokenizers 0.12.1