Well, I don’t quite understand your code. For example combined
is never used afterwards. It’s also not quite clear to me if all your sequences have the same length. In this case, you could of course do something like
self.linear1 = nn.Linear(seq_len*hidden_dim , 128)
If your sequences do not have the same length, but there is a maximum sequence length max_seq_len
, then you can try
self.linear1 = nn.Linear(max_seq_len*hidden_dim , 128)
and then pad all sequences that are shorter than max_seq_len
.