Question about train_forward_bce function in ManifoldModel

ExplainableML / czsl

PyTorch CZSL framework containing GQA, the open-world setting, and the CGE and CompCos methods.

GNU General Public License v3.0

111 stars 27 forks source link

Question about train_forward_bce function in ManifoldModel #6

Closed bighuang624 closed 3 years ago

bighuang624 commented 3 years ago

Hi,

Thanks for the high quality open-resourced codebase.

I suspect that the Line 59 of czsl/models/manifold_methods.py should be

composed_clf = self.compose(sampled_attrs, sampled_objs)

instead of the current

composed_clf = self.compose(attrs, objs)

As I am a beginner in the CZSL field, I apologize for any misunderstandings that may exist. Looking forward to any response.

mancinimassimiliano commented 3 years ago

Hi @bighuang624,

You are right, thanks for spotting that out. That part has been taken from this repo and affects only RedWine that has not been included as baseline in our experiments. We will fix it.

mancinimassimiliano commented 3 years ago

Fixed together with the embedding init strategy. I am closing this, thanks again @bighuang624!