search

thunlp / OpenKE

An Open-Source Package for Knowledge Embedding (KE)
3.83k stars 985 forks source link

Optimal hyper parameters for FB15K and WN18 for NEW OpenKE-Pytorch branch #242

Closed rlafraie closed 3 years ago

rlafraie commented 4 years ago

I just tried the hyper parameters from the FB15K example in the example file train_transe.py (PyTorch Old) which didn't work very well.

Does anyone know the optimal hyper parameters for the new OpenKE-Pytorch branch?

Many Thanks in advance!

LiuYuemei111 commented 3 years ago

The authors have provided the hyper-parameters of some models to achieve the state-of-the-art performace (Hits@10 (filter)) on FB15K237 and WN18RR. These scripts can be founded in the folder ./examples/