Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[QUESTION] Recreating Figure 9 in the paper #6

Open
CPA872 opened this issue Apr 5, 2024 · 4 comments
Open

[QUESTION] Recreating Figure 9 in the paper #6

CPA872 opened this issue Apr 5, 2024 · 4 comments
Labels

Comments

@CPA872
Copy link

CPA872 commented Apr 5, 2024

Your question
Dear Authors,

In Figure 9 of the paper, the Speedup is used as the x-axis. I am wondering what are the early-exit thresholds leading to the corresponding speedup? Can you shed light on this?

image

@pan-x-c
Copy link
Owner

pan-x-c commented Apr 5, 2024

In the experiment above, each line contains 21 points, and from left to right the threshold decreases by 0.05 each time from 1.0 to 0.0.

@CPA872
Copy link
Author

CPA872 commented Apr 6, 2024

Can you give a description of the evaluation environment? For example, what were the task prompts on the CNN/Dailymail dataset?

The provided checkpoint on ModelScope should be enough to create Figure 9. Is this correct?

Thank you for your time!

@pan-x-c
Copy link
Owner

pan-x-c commented Apr 6, 2024

We use an internally maintained version of HELM that could not be opened due to other work involved, but there are no changes to the prompts and the number of tokens compared to the official open-sourced version.

The provided checkpoint is enough to create Figure.9 with HELM.

Copy link

github-actions bot commented Jun 5, 2024

Marking as stale. No activity in 60 days.

@github-actions github-actions bot added the stale label Jun 5, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

2 participants