Control Room: A central system to track and stop model workloads across our research infrastructure

本文由「Control Room: A central system to track and stop model workloads across our research infrastructure」的公开内容整理，完整内容与后续更新请点击文末链接阅读原文。

Control RoomLoading…ShareAs our models get more powerful and the associated safety and security risks become bigger, it becomes more and more critical that we can detect potential issues, and quickly stop models in their tracks when needed. To solve this, we built Control Room.Control Room is a central system to track model workloads across our research infrastructure and stop them quickly when safety or security concerns arise. It connects every research inference workload, including agentic sampling for model training and evaluation, data processing, and more. This works together with our misalignment monitoring systems, and halts model output within 35 seconds of a decision to intervene to stop or pause a job. It has been built to keep working even when other systems fail, and to stop model output if Control Room itself goes down.Control Room is already deployed across our research infrastructure, where the need is greatest. Our external model-serving stack is already much more centralized by nature, with a well-developed operational capability to quickly take down model deployments if needed.How it worksControl Room works by integrating a few systems: the workload, the inference engine, and the centralized Control Room service.The workload is the job being run, such as a reinforcement learning experiment, an evaluation, a high-volume sampling or data processing job. We require every research workload that does agentic inference to be associated with an approved safety case.The inference engine, or just engine, is the model itself. You can think of the engine as the brain of the model, running directly on GPUs. A workload normally uses one or more models, which might be running on one or more different inference engines. The same inference engine might be used by more than one different workload.The Control Room service tracks jobs and controls whether they are allowed to continue generating model output. Engines check with the Control Room service before fulfilling a request. If a job has been stopped, the service denies the request, preventing further output.Control Room is integrated in the following way: When starting a new job, the workload code requests a new job_id from Control Room.Whenever the workload needs to sample tokens from a model, it makes a request to an inference engine running that model, passing the job_id. The engine requests permission from Control Room to fulfill that request, passing the job_id along to Control Room.Control Room checks whether that job has been terminated, and if so, it denies the inference request. The engine responds to the workload's inference request with a refusal, stopping further model output from being produced for that job.How Control Room worksResearcherSomething triggers a decision to stop the jobResearch job job_042Inference engineRuns the model on GPUsControl Roomjob_039, activejob_039job_040, activejob_040job_041, activejob_041job_042, activejob_042STEP 1A researcher decides to start a jobThe process starts with a researcher launching a job—for example, an experiment or an evaluation.The job starts upJob code (or workload code) is owned and run by researchers. Inference engines might be started as part of this job, or standalone engines which are not associated with any particular job might already be running.The job registers with Control RoomThe workload code requests a job ID, associating the job with an approved safety case. The job itself makes this request.Control Room returns a job IDControl Room records the job and returns its identifier. The workload will attach this same ID to its inference requests.The job requests model output from an inference engineThe request includes the job_id. This request is made directly from the workload code and does not pass through Control Room.The engine checks permissionBefore fulfilling the request, the inference engine asks Control Room whether this job is still authorized to run.Control Room allows the requestThe job is sti