Skip to content

Regarding the Choice of Calibration Set #13

@ShiningSord

Description

@ShiningSord

Dear authors,

I would like to express my deep appreciation for the significant contribution your work has made to our community. It has been an invaluable resource for me.

I have encountered a question regarding the calibration set you chose when trying to figure out the code. I hope you can offer some suggestions.

In your implementation, when pruning LLM with FLAP, the calibration set is chosen as WikiText-v2, as shown in this line.

If I understand correctly, comparing the PPL on WikiText-v2 with previous work might be misleading, as previous pruning methods were not trained on WikiText-v2. Could you please confirm this?

Can you specify the calibration set you choose when you report the result in your paper? I can not find this in the paper.

Thanks for your attention, and I hope to hear from you!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions