- 자연어처리를 이용한 관계 추출
- python train.py
- python train_pytorch.py
- ex) python train.py --seed 2021 --epochs 10 --model xlm-roberta-large
- python inference.py --model [model_path]
- ex) python inference.py --model xlm-roberta-large/checkpoint-2000
- crossentropy, f1loss, focalloss, label_smoothing_loss가 있음
- 데이터를 불러들여오고, tokenize하는 함수
- 데이터 전처리 함수
## 학습 단일모델 최고 성능 0.792
python train.py --epochs 10 --max_len 300 --model xlm-roberta-large --batch_size 16 --preprocess convert9 --tokenize tokenized_dataset1 --drop 0.5 --seed 2022
## inference
python inference.py --model xlm-roberta-large12-convert9-epoch10-linear-tokenized_dataset1-300-2022/checkpoint-3100 --token xlm-roberta-large --max_len 300 --name xlm-roberta-large12-convert9-epoch10-linear-tokenized_dataset1-300-2023 --preprocess convert9 --tokenize tokenized_dataset1 --seed 2022