{
  "context": "Evaluates whether the solution uses Luigi's resource-aware scheduling to bound the transcode stage while honoring caller-provided limits and worker counts. It checks that Luigi resource tokens are declared, configured, and enforced to produce the expected serialization and scaling behavior.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Resource mapping",
      "description": "Transcode task declares a Luigi resources mapping (e.g., resources = {'accelerator': 1}) so the scheduler tracks accelerator slots instead of using manual locks or semaphores.",
      "max_score": 30
    },
    {
      "name": "Scheduler limits",
      "description": "Caller-provided resource_limits are loaded into Luigi's [resources] configuration (via luigi.configuration.get_config().set or equivalent) before building the pipeline, ensuring the scheduler, not custom code, enforces available tokens.",
      "max_score": 20
    },
    {
      "name": "Worker wiring",
      "description": "Pipeline execution passes the requested worker_count to luigi.build (or luigi.run CLI) so Luigi's worker pool plus resource tokens govern parallelism; worker count is not hardcoded.",
      "max_score": 15
    },
    {
      "name": "Single-slot serialization",
      "description": "With resource_limits specifying one accelerator slot, Luigi scheduling keeps at most one transcode task running at a time (as seen in the returned timeline/runtime) without any user-managed concurrency primitives.",
      "max_score": 20
    },
    {
      "name": "Scaled concurrency",
      "description": "When the accelerator limit is increased (e.g., two slots), Luigi allows up to that many concurrent transcode tasks and the reported runtime improves accordingly, demonstrating reliance on Luigi's resource-aware scheduler.",
      "max_score": 15
    }
  ]
}