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

Performances as a function of the batch size #10

Closed
morsingher opened this issue Oct 26, 2022 · 1 comment
Closed

Performances as a function of the batch size #10

morsingher opened this issue Oct 26, 2022 · 1 comment

Comments

@morsingher
Copy link

Hi, thanks for releasing this cool work!
I have a question about Fig. 4 in your paper and the related paragraph. Since you train for 25k iterations with batch_size = 8, do you also increase (or decrease) the number of iterations when the batch size decreases (or increases, respectively), or is it always kept fixed at 25k?
Thank you in advance for the answer.

@aharley
Copy link
Owner

aharley commented Oct 26, 2022

Great question. For those low-batchsize models I had to train longer to give them a fair chance. (With lower batchsize, sometimes val is still increasing after 25k.) I think B2 and B4 had 50k, and B8 maybe 40k.

@aharley aharley closed this as completed Nov 4, 2022
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

2 participants