- π Competition Description
- πΎ Dataset Description
- π Folder Structure
- βοΈ Set up
- π» How to Run
κ΄κ³ μΆμΆ(Relation Extraction)μ λ¬Έμ₯μ λ¨μ΄(Entity)μ λν μμ±κ³Ό κ΄κ³λ₯Ό μμΈ‘νλ λ¬Έμ μ λλ€.
μ΄λ² λνμμλ λ¬Έμ₯, λ¨μ΄μ λν μ 보λ₯Ό ν΅ν΄ λ¬Έμ₯ μμμ λ¨μ΄ μ¬μ΄μ κ΄κ³λ₯Ό μΆλ‘ νλ λͺ¨λΈμ νμ΅μν΅λλ€. μ΄λ₯Ό ν΅ν΄ μ°λ¦¬μ μΈκ³΅μ§λ₯ λͺ¨λΈμ΄ λ¨μ΄λ€μ μμ±κ³Ό κ΄κ³λ₯Ό νμ νλ©° κ°λ μ νμ΅ν μ μμ΅λλ€.
Dataset | train | test |
---|---|---|
λ¬Έμ₯ μ | 32470 | 7765 |
λΉμ¨ | 80 | 20 |
-
id (λ¬Έμμ΄) : λ¬Έμ₯ κ³ μ ID
-
sentence (λ¬Έμμ΄) : μ£Όμ΄μ§ λ¬Έμ₯
-
subject_entity (λμ λ리) : 주체 entity
-
object_entity (λμ λ리) : κ°μ²΄ entity
-
label : (λ¬Έμμ΄) 30κ°μ§ labelμ ν΄λΉνλ 주체μ κ°μ²΄κ° κ΄κ³
-
source : (λ¬Έμμ΄) λ¬Έμ₯μ μΆμ²
-
wikipedia (μν€νΌλμ)
-
wikitree (μν€νΈλ¦¬)
-
policy_briefing (μ μ± λ³΄λ μλ£?)
-
βββπconfig
β βββ base_config.yaml
β βββ custom_config.yaml
β
βββπdata_loaders
β βββ data_loader.py β λ°μ΄ν°μ
μ λ‘λν©λλ€.
β βββ preprocessing.py
β
βββπdataset
β βββπdev
β β βββ dev.csv β dev(valid) λ°μ΄ν°
β βββπpredict
β β βββ predict.csv β μμΈ‘ν΄μΌνλ λ°μ΄ν°
β β βββ sample_submission.csv β μν λ°μ΄ν°
β βββπpretrain
β β βββ all_data.csv β train + test λ°μ΄ν°
β β βββ train.csv
β βββπtest
β β βββ test.csv β λͺ¨λΈ νμ΅ ν λ§μ§λ§ νκ°μμ μ¬μ©νλ λ°μ΄ν°
β βββπtrain
β βββ train.csv β νμ΅ λ°μ΄ν°
| βββ gpt_autmentation, roberta_augmentation, pororo_augmentation.csv
β
βββπmodel
β βββ auxiliary.py
β βββ entity_roberta.py
β βββ loss.py
β βββ lstm.py
β βββ metric.py
β βββ model.py
β βββ rbert.py
β βββ recent.py
β
βββπprediction
β βββ sample_submission.csv
β βββ submission.csv
β βββ submission_18-14-46.csv β inferenceνλ κ²½μ°, 'λ μ§-μκ°-λΆ.csv'κ° λ€μ λΆμ
β
βββπstep_saved_model β save_steps 쑰건μμ λͺ¨λΈμ΄ μ μ₯λλ κ²½λ‘.
β βββπklue-roberta-large β μ¬μ©ν λͺ¨λΈ
β βββπ18-14-42 β μ€νν λ μ§-μκ°-λΆ
β βββπcheckpoint-500 β μ μ₯λ 체ν¬ν¬μΈνΈ-μ€ν
β
βββπtrainer
β βββ trainer.py
β
βββπutils
β βββ util.py
β
βββ dict_label_to_num.pkl
βββ dict_num_to_label.pkl
βββ inference.py β inference μ½λ
β
βββ main.py β train.pyμ inference.py μ€ν μ½λ
β ex) trainνλ κ²½μ° β python main.py -mt
β inferenceνλ κ²½μ° β python main.py -mi
β
βββ tapt_pretrain.py β tapt task μ½λ
βββ train.py β train μ½λ
βββ train_ray.py β hyperparameter search μ½λ
βββ train_raybohb.py
$ pip install -r requirements.txt
train : dev : test = 8 : 1 : 1
$ python main.py -mt
$ python main.py -mi
$ python main.py -mtp