Inquiry About Dataset Construction for LinkGPT

[William3242]

I have been working on implementing your LinkGPT method for a specific downstream task in my field. I am excited to share that I have resolved the environment dependencies and successfully reproduced some of the results based on your GitHub repository.

Currently, I am preparing to adapt my own dataset to use with your method. But I noticed that the dataset shared on your GitHub repository is already preprocessed as.pklfiles. I would greatly appreciate it if you could provide any guidance, tutorials, or best practices for constructing and preprocessing datasets for LinkGPT.

Specifically, I am curious about:

1. The expected data format and any key requirements for compatibility with LinkGPT.
2. Steps or tools you recommend for preprocessing raw data into the provided .pkl format.
3. Any additional considerations or challenges I should be aware of during this process.

Thank you in advance for your time and assistance. Your research has been incredibly inspiring and helpful to my work, and I look forward to applying LinkGPT to my own dataset.