Parameters of HyperTuning

RUCAIBox / RecBole

A unified, comprehensive and efficient recommendation library

https://recbole.io/

MIT License

3.49k stars 616 forks source link

Parameters of HyperTuning #2032

Open AthinaKyriakou opened 8 months ago

AthinaKyriakou commented 8 months ago

I have been working with the HyperTuning from recbole.trainer.

On the provided documentation, I cannot find what are the parameters early_stop and max_evals and what is their relation to the parameter epochs specified in the configuration yaml file (example).

Could you clarify this?

Thanks a lot for your work and support on this project!

BishopLiu commented 3 months ago

Hi, max_evals refers to the maximum number of searches, and the hyper-param search stops when max_evals is reached. Additionally, the search will stop if a better parameter is not found after early_stop hyper-param combinations have been tried.