model-optimizer

NVIDIA's model optimization library providing quantization, pruning, distillation, NAS, and speculative decoding APIs to compress models for TensorRT-LLM, TensorRT, vLLM, and SGLang deployment.