def forward()

in src/speech_reps/models/decoar.py [0:0]

• 26 lines of code
• 3 McCabe index (conditional complexity)

    def forward(self, features, padding_mask=None):
        max_seq_len = features.shape[1]
        features = self.post_extract_proj(features)

        if padding_mask is not None:
            extra = padding_mask.size(1) % features.size(1)
            if extra > 0:
                padding_mask = padding_mask[:, :-extra]
            padding_mask = padding_mask.view(padding_mask.size(0), features.size(1), -1)
            padding_mask = padding_mask.all(-1)

        seq_lengths = (~padding_mask).sum(dim=-1).tolist()

        packed_rnn_inputs = pack_padded_sequence(features, seq_lengths,
                                                 batch_first=True,
                                                 enforce_sorted=False)

        packed_rnn_outputs, _ = self.forward_lstm(packed_rnn_inputs)
        x_forward, _ = pad_packed_sequence(packed_rnn_outputs,
                                   batch_first=True,
                                   total_length=max_seq_len)

        packed_rnn_inputs = pack_padded_sequence(self.flipBatch(features, seq_lengths), seq_lengths,
                                                 batch_first=True,
                                                 enforce_sorted=False)

        packed_rnn_outputs, _ = self.backward_lstm(packed_rnn_inputs)
        x_backward, _ = pad_packed_sequence(packed_rnn_outputs,
                                   batch_first=True,
                                   total_length=max_seq_len)
        x_backward = self.flipBatch(x_backward, seq_lengths)

        return torch.cat((x_forward, x_backward), dim=-1)