| from transformers import PretrainedConfig | |
| class SoundStreamConfig(PretrainedConfig): | |
| model_type = "soundstream" | |
| def __init__( | |
| self, | |
| channels: int = 32, | |
| latent_dim: int = 512, | |
| codebook_size: int = 1024, | |
| num_quantizers: int = 8, | |
| **kwargs, | |
| ): | |
| super().__init__(**kwargs) | |
| self.channels = channels | |
| self.latent_dim = latent_dim | |
| self.codebook_size = codebook_size | |
| self.num_quantizers = num_quantizers | |