How to choose pre-training data

[yhjiujiu]

hi, @acbull Thank you for your wonderful work. I have a few issues when using this method for my own dataset (10000 nodes). 1. How to select the pre-training data at this time, is it to randomly sample 90% of the data from the training data set? 2. My data set is in the form of triples (sub, rel, obj). At this time, the method of processing data still uses proprecess.py? ?

thanks!!

[Gender123]

@yhjiujiu ,hello, in my opinion, the method proposed in this paper is aimed at large-scale HINs, if your dataset is not so large, you can utilize the entire dataset or investigate your own pre-process method.