[dask] early_stopping_rounds for dask

[pseudotensor]

I'm confused about the latest changes for dask that support evaluation sets. But this is not really super useful unless we have early stopping. I see it's still not supported. When is it planned?

When I modified dask-lightgbm package, it was pretty easy to support both eval sets and early stopping, so I'm wondering why it's not here. I assume there is some sync issue?

h2oai/dask-lightgbm@8cc8e83

[jameslamb]

When is it planned?

If you'd are interested in contributing support for early stopping in LightGBM's Dask interface, comment on #3712 saying so and we'd be happy to review a pull request.

Otherwise, you can follow #3712 to be updated. At the moment, I'm not aware of any maintainers working on it. I personally am focused on R-package issues for the next minor release (#4310).

When I modified dask-lightgbm package, it was pretty easy to support both eval sets and early stopping,

The commit you've linked ignores significant details, such as how to handle the case where an eval set's partitioning results in some workers not getting any eval data. It also notably excludes documentation and tests.

I encourage you to read the significant discussion on #4101 and #3952 to understand what is challenging about adding early stopping to the Dask interface in this project.

[no-response]

This issue has been automatically closed because it has been awaiting a response for too long. When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one. Thank you for taking the time to improve LightGBM!

[github-actions]

This issue has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.