| """ | |
| Placeholder: real LLM code benchmark scripts are in the `jobs/` directory. | |
| - `jobs/run_real_llm_standalone.py` — v1 baseline with Qwen2.5-Coder-0.5B | |
| - `jobs/run_real_llm_standalone_v2.py` — v2 with chat templating fix | |
| These are self-contained GPU job scripts that inline the OCC components | |
| to avoid import issues in sandbox environments. | |
| """ | |