PyTorch Code for Adversarial and Contrastive AutoEncoder for Sequential Recommendation.
- python 3.6+
- PyTorch
- tqdm
- tensorboardX
- numpy
Run train.py
:
python3 train.py
The dataset is set to ml-1m
by default. You can change it by setting the hyper_params
in train.py
. For the convenience of reproduction, we provide 3 preprocessed datasets: ml-latest
, ml-1m
and ml-10m
. All of the lines in the datasets are formatted as [USER_ID] [ITEM_ID]
ordered by interaction timestamps.
If you want to train this model on your own datasets, you can save your preprocessed dataset files under datasets/
. You also need to add one item in dataset_info.json
, which contains the information of the count of users and items as well as the seq_len
to use in the model.