Is it possible to evaluate relation prediction in PyKeen

[LuisaWerner]

Hi,

I would like to train Knowledge Graph Embeddings with PyKEEN and I am interested in results on relation prediction. So far, I've only seen that right-side $(h,r,?)$ and left-side evaluation is possible $(?,r,t)$. I would like to evaluate how well the models perform on the task $(h,?,t)$, in other words predicting which relation type links two entities, namely $(head, ?, tail)$. Is there any code to do this in a straightforward manner in pykeen?

Thanks in advance for your help :-)

[cthoyt]

Been working on this for a while: #728

Unfortunately there are tons of other improvements that we want to make first to get this to work more elegantly so it's never been finished

[LuisaWerner]

Thanks for the information, I will have a look :-)