Skip to content

Latest commit

 

History

History
79 lines (61 loc) · 2.87 KB

README.md

File metadata and controls

79 lines (61 loc) · 2.87 KB
Raw

Relation Extraction Task

This is 🌏Earth Coding Lab's repository for the 2nd boostcamp AI Tech competition (2022.11.14 ~ 2022.12.01 19:00).

The competition is on sentence-level Relation Extraction.

  • Here is our Wrap-Up Report (link)
  • Here is our Presentation (link)

Contributors

Yong Woo Song Woo Bin Park Jin Myeong Ahn Yeong Cheol Kang Gang Hyeok Lee
Github Github Github Github Github

Yong Woo Song   :   Result Analysis • Paper Research • Data Augmentation • Model Implementation
Woo Bin Park   :   Loss Function Analysis • Model Ensemble • Model Implementation
Jin Myeong Ahn   :   Code Refactoring • Model Implementation
Yeong Cheol Kang   :   MLFlow Customization • Model Customization • Model Implementation
Gang Hyeok Lee   :   Paper Research • Data Cleaning • Hyperparameter Tuning • Model Implementation

Hardware Used

  • NVIDIA TELSA V100
  • Ubuntu 18.04

Hardware Used

  • NVIDIA TELSA V100
  • Ubuntu 18.04

📄 Guideline

Setup

Install all the prerequisites in one go.

make setup

Code formatting & Check lint

make style

Code Testing

make test

Training

python main.py

Inference

python main.py --do_train=False --do_inference
python main.py --do_train=False --do_inference

Run Dashboard

make dashboard

Then you can acess dashboard through your web browser. (If you use VScode, check issue)

Data

In this competition, KLUE dataset for relation extraction was used. It can be downloaded from the link below: https://aistages-prod-server-public.s3.amazonaws.com/app/Competitions/000075/data/dataset.tar.gz

EDA

EDA can be seen from eda_basics.ipynb file under eda folder or check our Wrap-Up Report.

Leaderboard

Micro F1 AUPRC Rank
Public 74.5240 76.6242 7th
Private 74.5271 81.1231 2nd