search

Tharun24 / MACH

Extreme Classification in Log Memory via Count-Min Sketch
46 stars 16 forks source link

Is pretrained model available for those three datasets? #3

Open Odaimoko opened 4 years ago

Tharun24 commented 4 years ago

For each dataset, we have an ensemble of models. The file sizes of each of them is substantial and hence we don't provide pre-trained models. However, I can provide models for the smaller wiki10-31K dataset on the cloud if you insist. May I know how you want to use them? For just evaluation or any other purpose?

Odaimoko commented 4 years ago

Wiki10-31k can be sufficient. I am working on a project to investigate how sota extreme classification methods perform on tail labels. It would be nice not to train models again :)