---
library_name: transformers
license: mit
datasets:
- kor_nli
- klue
language:
- ko
pipeline_tag: zero-shot-classification
---

kor-nli 및 klue-nli로 학습하였습니다.

test accuracy(KLUE-NLI): 0.889667

```
model_name = 'deliciouscat/kf-deberta-base-cross-nli'
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=3).to(device)

def encode(examples):
    return tokenizer(examples['premise'], examples['hypothesis'], truncation=True, padding='max_length', max_length=128)
```