MetaCAT with grid search

Does anyone have experience of using a grid search for hyperparameter tuning of a MetaCAT model?

I haven’t explored it but it certainly would be useful. Looking forward to a potential PR

Hi @Yoshi I’ve tested raytune at some point with population-based training, the parameters used currently are picked so that they work in most use-cases, but if you have a particular case then using any hyperparameter search algorithm should improve the performance. In the end, MetaCAT is just an LSTM.

Thanks, do you have any code you could share for using raytune with MetaCAT?

Unfortunately nothing, this was just me playing around some time ago, nothing proper.

Hey @Yoshi ! Did you get anywhere with this? :smiley: