Downloads · 30 days
13
12% of all-time downloads
hwp0725/roberta-korean-hanja-stdict-mlm
roberta-korean-hanja-stdict-mlm is a fill-mask model from hwp0725. Use it when you need the model to fill a missing word. It is set up for transformers. The card lists the license as cc-by-sa-4.0.
hwp0725/roberta-korean-hanja-stdict 베이스 모델을 한자 코퍼스로 MLM(Masked Language Modeling) 파인튜닝한 모델입니다.
Downloads · 30 days
13
12% of all-time downloads
All-time downloads
113
Public
Parameters
345M
4.1 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors1.4 GB · 100%
From the Hugging Face model README
hwp0725/roberta-korean-hanja-stdict 베이스 모델을 한자 코퍼스로 MLM(Masked Language Modeling) 파인튜닝한 모델입니다.
| 항목 | 값 |
|---|---|
| 베이스 모델 | hwp0725/roberta-korean-hanja-stdict |
| 파라미터 수 | 345.1M |
| 학습 방식 | MLM (Masked Language Modeling) |
| 에포크 | 3 |
| 모델 | 설명 |
|---|---|
| hwp0725/roberta-korean-hanja-stdict | vocab 확장 베이스 모델 (MLM 학습 전) |
| hwp0725/roberta-korean-hanja-stdict-mlm | MLM 파인튜닝 완료 모델 (현재 모델) |
from transformers import AutoTokenizer, AutoModelForMaskedLM, pipeline
tokenizer = AutoTokenizer.from_pretrained("hwp0725/roberta-korean-hanja-stdict-mlm")
model = AutoModelForMaskedLM.from_pretrained("hwp0725/roberta-korean-hanja-stdict-mlm")
fill_mask = pipeline("fill-mask", model=model, tokenizer=tokenizer)
result = fill_mask("孔子曰:學而時習之,不亦[MASK]乎")
print(result)
| 항목 | 값 |
|---|---|
| Learning rate | 2e-05 |
| Train batch size | 16 |
| Seed | 42 |
| Optimizer | AdamW (betas=(0.9,0.999), epsilon=1e-08) |
| LR scheduler | linear (warmup 500 steps) |
| Epochs | 3 |
| Mixed precision | Native AMP |
CC BY-SA 4.0