YouTube Dataset

To download the dataset, please refer to google drive.

CreAgent

This contains the codes of the simulation platform equipped with CreAgent, which is used for long-term recommender system evaluation.

First, collected the dataset and put them into LLaMA-Factory/data by creating a new folder, e.g., LLaMA-Factory/data/youtube. The dataset should contain both creator and user. You can directly use the Small YouTube Dataset provided in the google drive.

The dataset format of provider is

[
    {
        "channel_name": "xxx",
        "history_categories": {
            "Entertainment": 98,
            "Film & Animation": 1,
            "Education": 1
        },
        "creation_frequency": xx,
        "history_items": []
      },
     ...
]

The dataset format of user is

[
    {
        "user_name": "xxx",
        "interest": [
            "People & Blogs",
            "Entertainment",
            "Film & Animation"
        ],
        "history": [
            "SS1ac8mAhhE",
            "KB8-cAU8kW4",
            ...
        ]
        },
        ...
]

Second, please determine the setups of the simulation platform and modify the config file:

cd LLaMA-Factory/src/llamafactory/config/config.yaml

To run the simulator, please enter the llama-factory dictionary

cd LLaMA-Factory/src/llamafactory/simulator

Then, you can run the simulator

python simulator.py

To change the configure setting, you can enter the LLaMA-Factory/src/llamafactory/config/config.yaml file and edit.