Closed shaoyx closed 6 years ago
Hi @simon0227
Thank you for reporting. It seems that it's my mistake. I'll check and fix my code.
This is due to the incorrect score calculation and I fixed it. Please retry to run TransE. it should work and converge the training loss.
It works now. Thanks~
Hi all, I ran TransE on WordNet with following command:![image](https://user-images.githubusercontent.com/2384668/34995154-9b8c7126-fb10-11e7-8f16-ea2e7a15ea02.png)
python train.py --mode pairwise --ent ../dat/wordnet-mlj12/train.entlist --rel ../dat/wordnet-mlj12/train.rellist --train ../dat/wordnet-mlj12/wordnet-mlj12-train.txt --valid ../dat/wordnet-mlj12/wordnet-mlj12-valid.txt --method transe
But the loss is not converging.Do I have set the wrong parameters?