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

restoreBestWeights support request. #8488

Open
PLtier opened this issue Jan 8, 2025 · 0 comments
Open

restoreBestWeights support request. #8488

PLtier opened this issue Jan 8, 2025 · 0 comments
Labels
type:feature New feature or request

Comments

@PLtier
Copy link

PLtier commented Jan 8, 2025

System information

  • TensorFlow.js version (you are using): 4.22.0
  • Are you willing to contribute it (Yes/No): Yes, I almost sent the PR before I read you would issue first (though there is a TODO)

Describe the feature and the current behavior/state.

I’d like to implement earlyStopping's restoreBestWeights parameter equivalent to python TF version. It’s been denoted in code, but it’s been unsupported.

Will this change the current api? How?

It will almost no change it. There is already restoreBestWeights parameter denoted in code and documentation, it’s just unsupported.

Who will benefit with this feature?

Anyone who wants to prevent resource waste and overfitting. Although there has been no Github issue, there was a TODO in code so there must have been a will for future implementation. I personally wanted to implement it for my browser extension which uses tfjs so that it uses less resources when training.

Any Other info.

You can see my proposed change code here: diff PLtier/tfjs at restore-best-weights. In tfjs-layers it passes almost every test - per 107 I had a few timeouts on my 2019 Mac. I'd like very much to contribute.

@PLtier PLtier added the type:feature New feature or request label Jan 8, 2025
@PLtier PLtier changed the title restoreBestWeights feature request. restoreBestWeights support request. Jan 8, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
type:feature New feature or request
Projects
None yet
Development

No branches or pull requests

1 participant