Before training the model, we have split some of the train datasets due to the size of the dataset, so you need to merge them into the original train dataset using merge_data.py (set the paths).
Follow the steps below ⬇️
make dataset for training:
sh data_gen/generate.shtrain with command:
## for step 1. preference extraction
sh shell/train_preference.sh
## for step 2. profile construction
sh shell/train_user.sh
sh shell/train_item.sh
## for step 3. reasoning-enhanced rating prediction
sh shell/train_imdb.sh
sh shell/train_amazon-book.shmerge the model for inference with command:
sh shell/merge.shinference with command:
sh shell/test_imdb.sh
sh shell/test_amazon-book.shcheck the results with command:
python test_result_inspect.py