Calculate the start/end indices in character-level
python utility/preprocess.py <path_to_raw_data> <path_to_output_data>
Split data into train and valid set, output_path is the target folder
python utility/split_data.py <data_path> <output_path>
The data_path is a folder containing two preprocessed csv files named train.csv and valid.csv. Information about arguments is listed in main.py
python main.py \
--name <exp_name>
--mode train \
--data_path <path_to_data> \
--model <model_name> \
--epoch <num_epoch> \
--batch_size <batch_size> \
The data_path is a folder containing raw data of testing csv files name test.csv.
python main.py \
--name <exp_name>
--mode inference \
--data_path <path_to_data> \
--model <model_name> \
--ckpt <path_to_checkpoint> \
The checkpoints should be placed in <result_path>/subset_{i}/best.ckpt, where i is the index of subset.
python main.py \
--name <exp_name>
--mode ensemble \
--data_path <path_to_data> \
--model <model_name> \
--result_path <folder_to_ckpt> \