JesseStover's picture
Update README.md
935f519 verified
---
{}
---
The L2AI-dictionary model is fine-tuned checkpoint of [klue/bert-base](https://huggingface.co/klue/bert-base) for multiple choice, specifically for selecting the best dictionary definition of a given word in a sentence. Below is an example usage:
```python
import numpy as np
import torch
from transformers import AutoModelForMultipleChoice, AutoTokenizer
model_name = "JesseStover/L2AI-dictionary-klue-bert-base"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForMultipleChoice.from_pretrained(model_name)
model.to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))
prompts = "\"κ°•μ•„μ§€λŠ” λ½€μ†‘λ½€μ†‘ν•˜λ‹€.\"에 μžˆλŠ” \"강아지\"의 μ •μ˜λŠ” "
candidates = [
"\"(λͺ…사) 개의 μƒˆλΌ\"μ˜ˆμš”.",
"\"(λͺ…사) λΆ€λͺ¨λ‚˜ 할아버지, ν• λ¨Έλ‹ˆκ°€ μžμ‹μ΄λ‚˜ 손주λ₯Ό κ·€μ—¬μ›Œν•˜λ©΄μ„œ λΆ€λ₯΄λŠ” 말\"μ΄μ˜ˆμš”."
]
inputs = tokenizer(
[[prompt, candidate] for candidate in candidates],
return_tensors="pt",
padding=True
)
labels = torch.tensor(0).unsqueeze(0)
with torch.no_grad():
outputs = model(
**{k: v.unsqueeze(0) for k, v in inputs.items()}, labels=labels
)
print({i: float(x) for i, x in enumerate(outputs.logits.softmax(1)[0])})
```
Training data was procured under Creative Commons [CC BY-SA 2.0 KR DEED](https://creativecommons.org/licenses/by-sa/2.0/kr/) from the National Institute of Korean Language's [Basic Korean Dictionary](https://krdict.korean.go.kr) and [Standard Korean Dictionary](https://stdict.korean.go.kr/).