Method decode

examples/models/voxtral_tts/model.py:1271–1279

codes: (B, 1+acoustic_dim, T) -> (B, semantic_dim+acoustic_dim, T)

(
        self, codes: torch.Tensor, dtype: torch.dtype = torch.float32
    )

        self.acoustic_dim = config.acoustic_dim

    def decode(
        self, codes: torch.Tensor, dtype: torch.dtype = torch.float32
    ) -> torch.Tensor:
        """codes: (B, 1+acoustic_dim, T) -> (B, semantic_dim+acoustic_dim, T)"""
        semantic_codes = codes[:, :1, :]
        acoustic_codes = codes[:, 1:, :]
        sem_emb = self.semantic_codebook.decode(semantic_codes).to(dtype)
        aco_emb = self.acoustic_codebook.decode(acoustic_codes, dtype)
        return torch.cat([sem_emb, aco_emb], dim=1)


# ---------------------------------------------------------------------------