10 Sep 2023 Nvidia claims new software library doubles LLM inference speed on H100 GPU Open source TensorRT-LLM comes out next month, targets generative AI workloads