dasayan05
find a mistake in the code
Is there a typo in the code on line 73 of bezierae.py ? C = torch.cat([c_final[-1, 0], c_final[-1, 0]], 1) should be changed to C = torch.cat([c_final[-1, 0], c_final[-1, 1]], 1)
It's an old code and I no longer maintain this. But, looking at a glance, I think you are right. It seems I did not include the LSTM backward state in c_final for computing the latent. However, it's correctly done for h_final. So I believe it is still getting a lot of information about the LSTM's end state (from h; both forward and backward). Empirically, I think it wouldn't make much difference. It may only increase the performance a bit, if at all.
Since I no longer maintain this, you may correct it yourself or send a PR.
