NVIDIA Model Optimizer

Overview

NVIDIA Model Optimizer is an open-source library covering quantization, distillation, pruning, and speculative decoding. It targets production deployment stacks including TensorRT and vLLM, making it useful for reducing inference cost and latency.