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.

Sign up for GitHub

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

Jump to bottom

mini-batch training #642

Merged
merged 10 commits into from
Jun 15, 2024
Merged

mini-batch training #642

merged 10 commits into from
Jun 15, 2024

Conversation

brucefan1983
Copy link
Owner

@brucefan1983 brucefan1983 commented Jun 13, 2024 •
edited
Loading

This PR aims to improve the training results when using mini-batches, based on the following tricks:

  • Increase the default regularization by a factor of $\sqrt{10}$.
  • Set up an upper limit of the learning rate: $\sigma_{\rm max}=0.01$, which is also the initial learning rate $\sigma_0$.
  • Prepare the mini-batches according to energy/atom values, maximizing the diversity for each batch.

@brucefan1983 brucefan1983 changed the title try sigma decay sigma_max and sigma_min Jun 15, 2024
@brucefan1983 brucefan1983 changed the title sigma_max and sigma_min mini-batch training Jun 15, 2024
@brucefan1983 brucefan1983 marked this pull request as ready for review June 15, 2024 10:28
@brucefan1983 brucefan1983 merged commit db325ae into master Jun 15, 2024
@brucefan1983 brucefan1983 deleted the sigma_decay branch June 15, 2024 15:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@shdchen shdchen shdchen approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@brucefan1983 @shdchen