{ "phone_former": "phone_former", "temp_former": "temp_former", "dep_former": "dep_former_csm", "phone_vocab_size": 73, "audio_vocab_size": 2049, "embedding_dim": 1024, "spk_embedding_dim": 192, "num_codebooks": 12, "num_phone_states": 4, "amortization_divisor": 16, "look_ahead": 2, "audio_window_size": 250, "phone_window_size": 350 }