evan11401/cs_ioc5008_0856043_finalproject

CS_IOC5008_0856043+0845051_FinalProject

Environment Setting

cuda 10.0(The latest version is cuda 10.1, so maybe you need to downgrade).

use nvcc --version to check

pip install imgaug
pip install lightgbm
pip install tqdm
#for albu
pip install torch==0.3.1
pip install torchvision==0.2.1
pip install scipy==1.1.0
pip install pandas
pip install tensorboardX
#for victor
pip install tensorflow-gpu==1.15.0
pip install Keras==2.2.4
pip install Keras-Applications==1.0.8
pip install opencv-python==3.4.2.16 
pip install pandas
pip install scikit-learn==0.19

Run the Code

Train

cd albu/src
./train_all.sh
./tune_all.sh
./predict_oof.sh

Predict

cd albu/src
./predict_test.sh
python merge_oof.py
python merge_test.py

Train GBM (LightGBM)

cd vicor
python train_classifier_lgbm.py
python create_submissions_lgbm.py --data_num 1 --test_fold merged_test
python create_submissions_lgbm.py --data_num 2 --test_fold merged_test

Train GBM (XGBoost)

cd vicor
python train_classifier_xgboost.py
python create_submissions_xgboost.py --data_num 1 --test_fold merged_test
python create_submissions_xgboost.py --data_num 2 --test_fold merged_test