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

Memory requirements + baseline configs #14

Open
fabian-sp opened this issue Jun 11, 2024 · 0 comments
Open

Memory requirements + baseline configs #14

fabian-sp opened this issue Jun 11, 2024 · 0 comments

Comments

@fabian-sp
Copy link

Hi,

first of all many thanks for making this repository accessible, great job!

I have two questions/comments:

  • it would be nice if there was an indication how much memory is roughly needed to load/prepare the datasets. I started with your example of SlimPajama-6B, and had some starting problems because I ran out of memory. Increasing the number of CPUs did the job, but it was a bit hard for me to find out how much is actually needed (and then you have to redo the download every time)
  • it would be also super helpful to have some benchmarks if you have them available: for example, for a given model and dataset, what is the best train/val loss you reached so far, and what is the optimizer setting that reached it. I don't know if you did many runs yourself, but if you have these information, it would be awesome to make an overview such that everyone can config a good baseline without having to do the tuning.

Thanks, and kind regards,
Fabian

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

1 participant