How to use from the
Use from the
Transformers library
# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("token-classification", model="Hyoungjun-yk/legal-instrument-doro-ner-finetuned")
# Load model directly
from transformers import AutoTokenizer, AutoModelForTokenClassification

tokenizer = AutoTokenizer.from_pretrained("Hyoungjun-yk/legal-instrument-doro-ner-finetuned")
model = AutoModelForTokenClassification.from_pretrained("Hyoungjun-yk/legal-instrument-doro-ner-finetuned", device_map="auto")
Quick Links

legal-instrument-doro-ner-finetuned

legal_instrument_generation ๋Ÿฐํƒ€์ž„์—์„œ ์‚ฌ์šฉํ•˜๋Š” ํ•œ๊ตญ์–ด ์ฃผ์†Œ/์œ„์น˜ ๋น„์‹๋ณ„ํ™” NER ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค.

์šฉ๋„

๋ฒ•๋ฅ  ๋ฌธ์„œ์— ํฌํ•จ๋œ ์ฃผ์†Œ์„ฑ ํ‘œํ˜„์„ ํƒœ๊น…ํ•ด ๋น„์‹๋ณ„ํ™”์— ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.

๋Œ€ํ‘œ label ์˜ˆ์‹œ๋Š” ์•„๋ž˜์™€ ๊ฐ™์Šต๋‹ˆ๋‹ค.

  • B-LOC_DO, B-LOC_SI, B-LOC_GU, B-LOC_DONG
  • B-ROAD, B-ROAD_NO
  • B-BLD_NAME, B-BLD_NAME_DONG, B-BLD_NAME_HO

ํŒจํ‚ค์ง•๋œ ์„ค์ • ๊ธฐ์ค€ label ์ˆ˜๋Š” 20๊ฐœ์ž…๋‹ˆ๋‹ค.

๊ธฐ๋ฐ˜ ๋ชจ๋ธ / ๊ตฌ์กฐ

  • architecture: BertForTokenClassification
  • hidden size: 768
  • max position embeddings: 512
  • vocab size: 32000

ํ•™์Šต ํŒŒ๋ผ๋ฏธํ„ฐ

  • learning rate: 5e-5
  • epochs: 10
  • train batch size: 8
  • eval batch size: 8
  • weight decay: 0.01
  • scheduler: linear
  • gradient accumulation: 1
  • logging steps: 20
  • seed: 42
  • fp16: False
  • bf16: False

๋Ÿฐํƒ€์ž„ ์‚ฌ์šฉ ์œ„์น˜

์ด ๋ชจ๋ธ์€ ์•„๋ž˜ ๊ฒฝ๋กœ๋กœ ๋‚ด๋ ค๋ฐ›์•„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.

  • local path: models/anonymization/doro_ner_finetuned

์„œ๋ฒ„ ์ฝ”๋“œ์—์„œ๋Š” legal-ai-server/app/anonymization/service.py์—์„œ ๋กœ๋“œํ•ฉ๋‹ˆ๋‹ค.

์ฐธ๊ณ 

์ด ๋ฆฌํฌ๋Š” ๋Ÿฐํƒ€์ž„ ์ฒดํฌํฌ์ธํŠธ ๋ฐฐํฌ ๋ชฉ์ ์ž…๋‹ˆ๋‹ค.
ํ•™์Šต ๋ฐ์ดํ„ฐ์…‹๊ณผ ์ „์ฒด ์‹คํ—˜ ๋กœ๊ทธ๋Š” ํฌํ•จํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.

Downloads last month
3
Safetensors
Model size
0.1B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support