printf

Reference Codes - printf: Preference Modeling Based on User Reviews with Item Images and Textual Information via Graph Learning

0. Create a virtual environment & Set up dependencies

python3 -m venv venv
source venv/bin/activate 
pip install -r requirements.txt

1. Download Amazon Data

cd gcn
bash ./download_amazon.sh 

2. Preprocess data & Generate review embeddings

python3 new_preprocess.py

3. Download corresponding images

python3 download_image.py

4. Fine tune multi-modality encoder & Generate item embeddings (CMIM)

cd albef
bash run_fine_tune.sh

P.S. Remember to download pretrained weights 4M/14M from ALBEF Repo.

5. Generate user embeddings (RAUM)

cd gcn
python3 -W ignore gen_user_embedding.py 

6. Train printf (EPIM)

python3 train.py 

7. Test printf

python3 test.py

P.S. We have set up all default values for the arg parse, please feel free to change each argument and run different experiments and datasets.