Method _tokenize_segment

generator.py:98–106

Returns: (seq_len, 33), (seq_len, 33)

(self, segment: Segment)

        return torch.cat(frame_tokens, dim=0), torch.cat(frame_masks, dim=0)

    def _tokenize_segment(self, segment: Segment) -> Tuple[torch.Tensor, torch.Tensor]:
        """
        Returns:
            (seq_len, 33), (seq_len, 33)
        """
        text_tokens, text_masks = self._tokenize_text_segment(segment.text, segment.speaker)
        audio_tokens, audio_masks = self._tokenize_audio(segment.audio)

        return torch.cat([text_tokens, audio_tokens], dim=0), torch.cat([text_masks, audio_masks], dim=0)

    @torch.inference_mode()
    def generate(