#TensorRT

También: tensorrt-llm

NVIDIA inference optimiser and runtime. Used to compile and quantise models, including TensorRT-LLM for large language models, to cut latency and raise GPU throughput.