Requirements:

Python==3.9 | PyTorch==1.12.1 | TorchData==0.4.1 | FreeRec==0.6.3

Use the following commands to prepare the environment (CUDA: 11.3):

conda create --name=PyT12 python=3.9; conda activate PyT12; bash setup_env.sh

Usage

This repo separately provides the implementaion of each conventional method mentioned in the paper. They can be conducted as follows

• CE:

python main.py --config=configs/AmazonBeauty_550_Chron.yaml --loss=CE

• BCE:

python main.py --config=configs/AmazonBeauty_550_Chron.yaml --loss=BCE --num-negs=1 --neg-pool=unseen

• BPR:

python main.py --config=configs/AmazonBeauty_550_Chron.yaml --loss=BPR --num-negs=1 --neg-pool=unseen

• SSM:

python main.py --config=configs/AmazonBeauty_550_Chron.yaml --loss=SCE --alpha=1 --neg-pool=all

• SCE:

python main.py --config=configs/AmazonBeauty_550_Chron.yaml --loss=SCE --alpha=100 --neg-pool=all

Note: Refer to benchmarks for logs of 5 independent runs.