- model.safetensors VERBATIM (bf16, sha256-checked against upstream LFS metadata)
- tokenizer files bundled for fully-offline runtime
- LICENSE.md: the Llama 3.2 Community License, as required for redistribution
- README carries the mandatory 'Built with Llama' notice
- structural verify (config + safetensors header: 183286-token vocab, bf16) before upload
Loaded by MAESTRO via backend/ai/models/midi_llm.py.
