| | |
|---|
| | | loss = self.ctc_weight * loss_ctc + (1 - self.ctc_weight) * loss_att + loss_pre * self.predictor_weight |
|---|
| | | |
|---|
| | | if self.use_1st_decoder_loss and pre_loss_att is not None: |
|---|
| | | loss = loss + pre_loss_att |
|---|
| | | loss = loss + (1 - self.ctc_weight) * pre_loss_att |
|---|
| | | |
|---|
| | | # Collect Attn branch stats |
|---|
| | | stats["loss_att"] = loss_att.detach() if loss_att is not None else None |
|---|
| | |
|---|
| | | mask_chunk_predictor=mask_chunk_predictor, |
|---|
| | | target_label_length=None, |
|---|
| | | ) |
|---|
| | | predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(pre_alphas[:, :-1], |
|---|
| | | encoder_out_lens) |
|---|
| | | predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(pre_alphas, |
|---|
| | | encoder_out_lens+1 if self.predictor.tail_threshold > 0.0 else encoder_out_lens) |
|---|
| | | |
|---|
| | | scama_mask = None |
|---|
| | | if self.encoder.overlap_chunk_cls is not None and self.decoder_attention_chunk_type == 'chunk': |
|---|
