1e15171f195d8094055eb8f460953fd6a840d9aa,cube/models/vocoder.py,BeeCoder,learn,#BeeCoder#,118

Before Change

            pred_output = self._predict_one(mgc[mgc_index], runtime=False)

            if mgc_index != len(mgc) - 1:
                losses.append(dy.l1_distance(pred_output, dy.inputVector(
                    wave[mgc_index * self.UPSAMPLE_COUNT:mgc_index * self.UPSAMPLE_COUNT + self.UPSAMPLE_COUNT])))

            if len(losses) >= batch_size:
                loss = dy.esum(losses)

After Change

                                                gs_output=wave[
                                                          mgc_index * self.UPSAMPLE_COUNT:mgc_index * self.UPSAMPLE_COUNT +
                                                                                          self.UPSAMPLE_COUNT])
                target_vec_1 = wave[
                               mgc_index * self.UPSAMPLE_COUNT:mgc_index * self.UPSAMPLE_COUNT + self.UPSAMPLE_COUNT]
                losses.append(dy.l1_distance(pred_output, dy.inputVector(target_vec_1)))

                history = wave[
                          (mgc_index + 1) * self.UPSAMPLE_COUNT - self.HISTORY:(mgc_index + 1) * self.UPSAMPLE_COUNT]

Instances