LLM4CDSR: Large Language Models Can Be a Good Bridge for Cross-domain Sequential Recommendation

This is the implementation of the paper "LLM4CDSR: Large Language Models Can Be a Good Bridge for Cross-domain Sequential Recommendation".

Configure the environment

To ease the configuration of the environment, I list versions of my hardware and software equipments:

• Hardware:
  • GPU: RTX 3090 24GB
  • Cuda: 12.0
  • Driver version: 525.105.17
  • CPU: AMD EPYC 7543 32-Core
• Software:
  • Python: 3.9.16
  • Pytorch: 2.3.1

You can pip install the requirements.txt to configure the environment.

Preprocess the dataset

You can preprocess the dataset and get the LLMs embedding according to the following steps:

1. The raw dataset downloaded from website should be put into /data/<amazon/elec/douban>/raw/. The Douban dataset can be obtained from this Download Link. The amazon and elec datasets can be obtained from this Download Link.
2. Conduct the preprocessing code data/<amazon/elec/douban>/handle.ipynb to get the data for cross-domain sequential recommendation. After the procedure, you will get the id file /data/<amazon/elec/douban>/hdanled/id_map.json and the interaction file /data/<amazon/elec/douban>/handled/<cloth-sport/elec-phone/book-movie>.pkl.
3. To get the LLMs embedding for cross-domain items, please run the jupyter notebooks /data/<amazon/elec/douban>/item_prompt.ipynb. After the running, you will get the LLMs item embedding file /data/<<amazon/elec/douban>/handled/itm_emb_np.pkl. Then, run the /data/<amazon/elec/douban>/pca.ipynb to get their dimension-reduced unified LLM embedding file itm_emb_np_all.pkl and local embedding itm_emb_np_<A/B>_pca128.pkl.
4. To get the user's comprehensive profiles and correpsonding LLM embeddings, run the python file /data/<amazon/elec/douban>/user_profile.py, and then you can get usr_profile_emb.pkl.

In conclusion, the prerequisite files to run the code are as follows: <cloth-sport/elec-phone/book-movie>.pkl, itm_emb_np_all.pkl, itm_emb_np_<A/B>_pca128.pkl, usr_profile_emb.pkl and id_map.json.

⭐️ To ease the reproducibility of our paper, we also upload all preprocessed files to this link.

Run and test

1. You can reproduce all LLM4CDSR experiments by running the bash as follows:

bash experiments/<amazon/elec/douban>/llm4cdsr.bash

2. The log and results will be saved in the folder log/. The checkpoint will be saved in the folder saved/.