skilld
All skills
nvidia avatar

/cuopt-numerical-optimization-api

@e0cd22d
by NVIDIA Corporationnvidia/skills3.5k stars
424

LP, MILP, and QP (beta) with cuOpt — Python, C, and CLI. Use when the user is solving LP, MILP, or QP with any cuOpt interface.

Ask your Agent

Use this Skill: https://skilld.dev/gh/nvidia/skills/cuopt-numerical-optimization-api

This session only. Nothing lands on disk.

assetspythonmps_solverresults.md

≈631 tokens on demand. Your agent reads this file only when SKILL.md points to it.

MPS Solver Results

Problem: air05.mps (MIPLIB benchmark)

Description: Airline crew scheduling - set partitioning problem

Problem Characteristics

  • Variables: 7195 (all binary)
  • Constraints: 426
  • Nonzeros: 52121
  • Best Known Optimal: 26374

Gap Tolerance Comparison

Comparing different MIP relative gap tolerances to show trade-off between solution quality and solve time.

Run Configuration

  • Time Limit: 60 seconds
  • cuOpt Version: 26.2.0
  • Device: Quadro RTX 8000 (47.24 GiB VRAM)
  • CPU: AMD Ryzen Threadripper PRO 3975WX (32 cores)

Results Summary

Gap Tolerance Objective Gap to Optimal Solve Time Nodes Explored
0.1% 26374 0.00% 8.42s 386
1.0% 26491 0.44% 3.23s 328

Key Observations

  1. Tighter gap finds optimal: The 0.1% gap tolerance found the exact best-known optimal solution (26374)
  2. Trade-off: The looser 1.0% gap converged faster (3.2s vs 8.4s) but with 0.44% suboptimality
  3. Both are fast: cuOpt solved this 7195-variable MILP in under 10 seconds

Detailed Solver Output (0.1% gap)

Solving a problem with 426 constraints, 7195 variables (7195 integers), and 52121 nonzeros

Presolve removed: 90 constraints, 1116 variables, 16171 nonzeros
Presolved problem: 336 constraints, 6079 variables, 35950 nonzeros

Root relaxation objective +2.58776093e+04

Strong branching using 7 threads and 222 fractional variables
Explored 386 nodes in 7.73s.

Optimal solution found within relative MIP gap tolerance (1.0e-03)
Solution objective: 26374.000000
relative_mip_gap 0.000992
total_solve_time 8.421934

Detailed Solver Output (1.0% gap)

Solving a problem with 426 constraints, 7195 variables (7195 integers), and 52121 nonzeros

Presolve removed: 90 constraints, 1116 variables, 16171 nonzeros
Presolved problem: 336 constraints, 6079 variables, 35950 nonzeros

Root relaxation objective +2.58776093e+04

Strong branching using 63 threads and 222 fractional variables
Explored 328 nodes in 1.09s.

Optimal solution found within relative MIP gap tolerance (1.0e-02)
Solution objective: 26491.000000
relative_mip_gap 0.009669
total_solve_time 3.233650

Usage

# Default: download air05.mps and solve with comparison
python model.py --compare --time-limit 60

# Solve custom MPS file
python model.py --file path/to/problem.mps --time-limit 300 --mip-gap 0.001

Source: SKILL.md on GitHub

Third-party checks

1 warning1mo3 checks · Risk SAFE
  • Gen Agent Trust Hub1mo

    The skill provides a comprehensive environment for modeling and solving optimization problems using the NVIDIA cuOpt library. It includes well-documented examples and reference implementations for Python, C, and CLI interfaces. Security analysis identified a remote data download from a reputable academic source and standard data ingestion patterns for optimization files, neither of which present malicious risk.

  • Socket1mo

    No alerts

  • Snyk1mo

    Risk: MEDIUM · 1 issue

Provenance

Signed by skilld at e0cd22d. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 2 months ago

Capability

version
26.10.00
Other metadata
metadata
{
  "author": "NVIDIA cuOpt Team",
  "tags": [
    "cuopt",
    "linear-programming",
    "milp",
    "qp",
    "python",
    "c-api",
    "cli"
  ]
}

README badge

README badge for nvidia/skills/cuopt-numerical-optimization-api