How would you describe them?
High-performance single-GPU inference for selected model checkpoints and GPUs.
