Practical and Reproducible Symbolic Music Generation by Large Language Models with Structural Embeddings | Synapse