Introduction
This is the Pytorch implementaion of some nowledge graph embedding(KGE) models. And I have test these models with popular datasets.
Implemented features
Models:
- TransE
- TransH
- TransD
- TransA
- ConvE
- SimplE
- RotatE
Usage
- Download datasets and put it in
./data
directory. - Then you can run commend as follows to train/test/valid the models. All training codes can be found in
./examples/
python3 ./examples trainTransE.py
If you want to modify parameters or datasets. You can rewrite training files and overwrite raw parameters. The parameters at Config.py
come from their paper.
Info
- TransA model has a question about memory explosion
Links
Datasets: