search

thunlp / FewRel

A Large-Scale Few-Shot Relation Extraction Dataset
https://thunlp.github.io/fewrel.html
MIT License
727 stars 165 forks source link

There is something wrong in GNN and MetaNet model #33

Closed githwd2016 closed 4 years ago

githwd2016 commented 4 years ago

When choose gnn or metanet, the parameter in forward is Q that means Num of instances for each class in the queryset. However, in framework, it passed Q N_for_train + na_rate Q that means total_Q.

gaotianyu1350 commented 4 years ago

The bug has been fixed.

icoderzqliu commented 2 years ago

The bug has been fixed.

@githwd2016 githwd2016 commented on 13 Mar 2020

The bug has been fixed.

I'm sorry, it seems that this issue isn't sloved. In metanet:

def forward(self, support, query, N, K, Q):

In framework:

logits, pred = model(support, query, N_for_train, K, Q * N_for_train + na_rate * Q)
icoderzqliu commented 2 years ago

When choose gnn or metanet, the parameter in forward is Q that means Num of instances for each class in the queryset. However, in framework, it passed Q N_for_train + na_rate Q that means total_Q.

Is this issue truly solved? Why the code is still the same?