针对特定模型检查点和 GPU 的高性能单 GPU 推理能力。
High-performance single-GPU inference for selected model checkpoints and GPUs.
