This is my implementation for the paper：
Rose Catherine, William Cohen. "TransNets: Learning to Transform for Recommendation." (2017).
- python 3.8
- pytorch 1.70
data/Digital_Music_5.json.gz:
Download from http://jmcauley.ucsd.edu/data/amazon (Choose Digital Music)
- Preprocess origin dataset in json format to train.csv,valid.csv and test.csv.
python preprocess.py
Train and evaluate the model
- TransNet
python main.py
- TransNet-Ext
Arguments --user_count and --item_count are necessary
and you can get their values from output of python preprocess.py
python main.py --extension True --user_count 5541 --item_count 3568