schelotto / Neural_Speed_Reading_via_Skim-RNN_PyTorch

PyTorch implementation of "Neural Speed Reading via Skim-RNN"
MIT License
18 stars 9 forks source link

Typo in copying of cell-state? #9

Open leox1v opened 4 years ago

leox1v commented 4 years ago

c_tilde = torch.transpose(torch.stack( [c_l_, torch.cat([c_s[:, :self.small_cell_size], c_l_[:, self.small_cell_size:self.large_cell_size]], dim=1) ], dim=2), 1, 2)

Shouldn't it be c_l[:, ...] here instead of c_l_[:, ...]? The way it is right now it seems like you're using the large cell every time. Or am I missing something?