We reuse the implementation of RAGatouille.
Install RAGatouille with pip:
pip install ragatouille
Firstly, prepare a number of data folders named fold_i, each data folder includes three files: train.json, valid.json, test.json. Each file is a JSON file, where each JSON object is formatted as: {qdpair_id: {dataset_id: rel_score, ...}, ...}.
Next, prepare query_info.json and dataset_info.json, where each JSON object is formatted as: {id: info_text, ...}.
Then, run commands as follow:
CUDA_VISIBLE_DEVICES=0,1,2,3 \
python grid_search.py \
--doc_maxlen 512 \
--top_k 20 \
--fold_start 0 \
--fold_end 5 \
--learning_rate 1e-05 \
--batch_size 8 \
--mode train