The author's claims are that some settings included in the current Recommenders version, beyond the model part, are not strictly following the original paper.
Usually for research exploration, people’s goal is to explore which model structure is better, while leaving other training settings unified for fair comparison.
It is doable that we also add all the training tricks listed in the table into Recommenders, so that not only GRU4Rec, but other models can use them.
For now, we will remove all references to GRU4Rec to avoid misunderstandings.
Description
Following the concerns of https://arxiv.org/abs/2307.14956, we removed the references to GRU4Rec.
The author's claims are that some settings included in the current Recommenders version, beyond the model part, are not strictly following the original paper.
Usually for research exploration, people’s goal is to explore which model structure is better, while leaving other training settings unified for fair comparison.
It is doable that we also add all the training tricks listed in the table into Recommenders, so that not only GRU4Rec, but other models can use them.
For now, we will remove all references to GRU4Rec to avoid misunderstandings.
We welcome the original authors or other people in the open source community to contribute the original version of GRU4Rec https://github.com/hidasib/GRU4Rec_PyTorch_Official
Related Issues
References
Checklist:
staging branchand not tomain branch.