LLM-EDT: Large Language Model Enhanced Cross-domain Sequential Recommendation with Dual-phase Training

This is the implementation of the paper "LLM-EDT: Large Language Model Enhanced Cross-domain Sequential Recommendation with Dual-phase Training".

Configure the environment

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

• Hardware:
  • GPU: RTX 4090 24GB
  • Cuda: 12.0
• Software:
  • Python: 3.11.16
  • Pytorch: 2.4.1

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 the website should be put into /data/<amazon/elec/douban>/raw/. Specifically, The Cloth-Sport and Electronics - Cell Phone and Food-Kitchen datasets can be obtained from this Download Link.
2. Conduct the preprocessing ipynb data/<amazon/elec/food>/ to process the data augmentation for cross-domain sequential recommendation. After the procedure, you will get the id file /data/<amazon/elec/douban>/handled/aug_id_map.json and the interaction file /data/<amazon/elec/douban>/handled/<aug_cloth-sport/aug_elec-phone/aug_food_kitchen>.pkl , the LLMs item embedding file /data/<<amazon/elec/douban>/handled/aug_itm_emb_np.pkl, and the unbiased user profile embedding: domain_split_usr_profile_emb.pkl.

In conclusion, the prerequisite files to run the code are as follows: aug_<cloth-sport/elec-phone/book-movie>.pkl, aug_itm_emb_np_all.pkl, domain_split_usr_profile_emb.pkl and aug_id_map.json.

⭐️ To ease the reproducibility of our paper, we also upload all preprocessed files to this link. The well-trained weight file can be found in this link. The log files can be found in this link.

Run and test

1. You can reproduce all LLM-EDT experiments by running the bash step by step as follows: Global Pretraining Stage:

bash experiments/<amazon/elec/food>/one4all.bash

Domain Fine-tuning Stage:

bash experiments/<amazon/elec/food>/domain_adapter.bash

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