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

Use Dadaptation for all DL models #327

Merged
merged 1 commit into from
Jun 26, 2024
Merged

Use Dadaptation for all DL models #327

merged 1 commit into from
Jun 26, 2024

Conversation

jakobnissen
Copy link
Member

Deprecate the learning rate.

Closes #267

@jakobnissen jakobnissen added the Needs benchmark Must benchmark before merging this label Jun 25, 2024
This optimizer is more efficient, and makes the learning rate obsolete, which
is nice from a UI standpoint.
@jakobnissen
Copy link
Member Author

Benchmarks show a 1% decrease in NCs, which is probably within measurement error. Also, the overall loss seems to get slightly better.
I'm going to merge this now anyway, because getting rid of the learning rate is a good idea. @sgalkina FYI, we might consider upgrading to prodigy, which is a newer, shinier optimiser made by the same people.

@jakobnissen jakobnissen removed Needs benchmark Must benchmark before merging this labels Jun 26, 2024
@jakobnissen jakobnissen merged commit b75f79d into master Jun 26, 2024
5 checks passed
@jakobnissen jakobnissen deleted the dadapt branch June 26, 2024 08:41
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

Successfully merging this pull request may close these issues.

Use D-Adaptation for all DL models
1 participant