| """Config for the standalone VocBulwark speaker encoder.""" | |
| from transformers import PretrainedConfig | |
| class SpeakerEncoderConfig(PretrainedConfig): | |
| model_type = "vocbulwark_speaker_encoder" | |
| def __init__(self, speaker_embed_config=None, embedding_size=None, | |
| raw_sample_rate=22050, **kwargs): | |
| super().__init__(**kwargs) | |
| # The wav2vec2-based encoder's own config (dict), plus the produced | |
| # embedding dimension and the sample rate its input audio is expected at. | |
| self.speaker_embed_config = speaker_embed_config | |
| self.embedding_size = embedding_size | |
| self.raw_sample_rate = raw_sample_rate | |