{ "model_type":"vits", "auto_map": { "AutoModel": "modeling_vits.ModVitsModel" }, "sampling_rate": 24000, "speaker_embedding_size": 256, "vocab_size": 1260, "speaker_names":["Assamese-Female", "Assamese-Male", "Bengali-Female", "Bengali-Male", "Bodo-Female", "Bodo-Male", "Dogri-Female", "Dogri-Male", "Kannada-Female", "Kannada-Male", "Maithili-Male", "Malayalam-Female", "Marathi-Female", "Marathi-Male", "Nepali-Female", "Punjabi-Female", "Punjabi-Male", "Sanskrit-Male", "Tamil-Female", "Telugu-Female"], "emotion_names":["ALEXA", "ANGER", "BB", "BOOK", "CONV", "DIGI", "DISGUST", "FEAR", "HAPPY", "NEWS", "SAD", "SURPRISE", "UMANG", "WIKI"], "undefined_emotion_index":[9,11,13] }