Controlled generation of OS LLMs – without impacting latency

TitanML Takeoff Inference Server demonstrating controlled generation