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

Cannot reproduce generated samples in UCLA-AGI/SPIN_iter0 #32

Open
StarDewXXX opened this issue May 22, 2024 · 3 comments
Open

Cannot reproduce generated samples in UCLA-AGI/SPIN_iter0 #32

StarDewXXX opened this issue May 22, 2024 · 3 comments

Comments

@StarDewXXX
Copy link

StarDewXXX commented May 22, 2024

I use revision="ac6e600eefcce74f5e8bae1035d4f66019e93190" to generate data for iter0, but cannot reproduce generated samples in UCLA-AGI/SPIN_iter0. Since do_sample is default to be false, the generation should be the same. Below is an example:

#########
How does the location of the Sydney Conservatorium of Music impact the academic and professional opportunities available to music students, and how does the conservatorium support student engagement with the music industry in Australia?

[reproduced with generate.py]: The location of the Sydney Conservatorium of Music in the heart of Sydney's central business district provides students with access to a wide range of academic and professional opportunities. The conservatorium is situated in the historic University of Sydney Quadrangle, ........(omitted)

[provided in UCLA-AGI/SPIN_iter0]: he Conservatorium is located in the heart of the city of Sydney, specifically in the central business district. This provides students with unique access to academic and professional opportunities within the city, including the opportunity to learn from top faculty members in the music ........(omitted)

@AGTSAAA
Copy link

AGTSAAA commented May 27, 2024

I also have the same problem. I can not reproduce the generated samples.

@angelahzyuan
Copy link
Collaborator

Have you considered that generation results might vary across different machines? Even with identical settings, the outcomes could differ depending on the GPU or the number of GPUs used. Are there any performance differences?

@angelahzyuan
Copy link
Collaborator

Also, our dataset posted on hf were generated using the hf generation, not vllm.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants