Below is the instruction to replicate the training, evaluation, and test processes. Note, some parts vary for the online-learning model.
- please use requirements.txt file with below command. recommend to create a virenv before installation.
step 1: conda create -n tasks python==3.7.3
step 2: conda install jupyter==1.0.0
step 3: conda install spyder==4.0.1
step 4: pip install -r requirement.txt
Note spyder and jupyter are included. But feel free to remove them if you plan not to create a blank virenv.
- If you want to use the provided trainset for retraining, please use file trainSet_enriched.csv.
- change cwd to the folder raw data is saved.
python scripts/train_pipeline.py outputs/data/trainSet_enriched.csv
- If you want to use the provided evalset for evaluation, please use file trainSet_eval.csv.
- change cwd to the folder raw data is saved.
python scripts/evaluation.py outputs/data/trainSet_eval.csv
- change cwd to the folder raw data is saved.
python scripts/prediction.py candidateTestSet.txt
If in need of changing some configuration of the model, visit config.py file. Please bear in mind that not all hyperparameters are exposed in config.py.
The raw data files are assumed and recommended to be saved in one level above the folder of the scripts.