NVIDIA cuOpt — математическая оптимизация с GPU

★ 7.6 · content

cuopt-numerical-optimization-api is a Claude Code skill that models and solves LP, MILP, and QP (beta) optimization problems using NVIDIA cuOpt's GPU-accelerated solver. It covers three interfaces — Python API, C/C++ API, and CLI for MPS files — and supports third-party modeling tools including AMPL, GAMS/GAMSPy, PuLP, JuMP, Pyomo, and CVXPY with near-zero code changes. Built-in guidance helps select the right problem type: LP for continuous flows and proportions, MILP for integer counts and yes/no decisions, QP for quadratic objectives such as variance minimization or least squares. The skill also explains dual value availability, common solver issues, and key settings like time limits and MIP gap tolerances. Version 26.10.00, Apache-2.0 license.