himkt comments

Results 26 comments of


                                            himkt

Improve typing coverage (4/n)

I want to try `pytorch_lightning/trainer/trainer.py`. 🚋

Improve typing coverage (4/n)

Thank you for the advice! OK, let me try `training_epoch_loop.py`!

Extend the type annotation of `catch` argument of `study.optimize`

> Wouldn't we do something like Optional[Union[Sequence[Type[Exception]], Type[Exception]]]? I think so too (`|` operator is available from Python 3.10 so we cannot use it yet). > Similarly for states: Optional[Sequence[optuna.trial._state.TrialState]...

Extend the type annotation of `catch` argument of `study.optimize`

@nzw0301 > I prefer relaxing this type annotation to Optional[Container[optuna.trial._state.TrialState]] Do you think it's better to allow other container objects such as `set`? It feels little bit too general to...

Extend the type annotation of `catch` argument of `study.optimize`

I see. Thank you for explaining. 📝

Extend the type annotation of `catch` argument of `study.optimize`

@BasLaa, thank you for improving `states`! We wait another PR to modify `catch` as well (ref. https://github.com/optuna/optuna/pull/3359#issuecomment-1060454209). Please feel free to mention if I can help you!

Extend the type annotation of `catch` argument of `study.optimize`

Hi @BasLaa, please tell me if I can help you.

How to make Sampler avoid a known-to-fail parameter configuration?

Hello @JRice15, is it possible to raise `TrialPruned` instead making the trial fail? AFAIK, failed trials are not used in samplers to pick a new parameter configuration. Catching the Out-Of-Memory...

Update CLI tutorial

@hppRC is now working on this topic. 👍

Update CLI tutorial

Hi @hppRC, please feel free to tell me if I can help you. :pray: