--- base_model: - SparkAudio/Spark-TTS-0.5B --- Quantized Qwen2.5-0.5B-Instruct from Spark-TTS.