| # Launch the orchestrator process using Ray (or as a standalone async task) | |
| # Example usage: | |
| # scripts/launch_orchestrator.sh <model_path> | |
| MODEL_PATH=$1 | |
| if [ -z "$MODEL_PATH" ]; then | |
| echo "Usage: $0 <model_path>" | |
| exit 1 | |
| fi | |
| # Ensure Ray is running | |
| if ! ray status > /dev/null 2>&1; then | |
| ray start --head | |
| fi | |
| # Run orchestrator as a background Ray task | |
| ray exec "import src.orchestrator.scheduler as sch; from src.rollout.version_manager import VersionManager; vm = VersionManager(); async def sync(v): pass; orch = sch.Orchestrator(vm, sync); asyncio.run(orch.run())" --address=auto & | |
| echo "Orchestrator launched for model at $MODEL_PATH" | |