K-Dense-AI/scientific-agent-skills / skills/pymoo
pymoo
VerifiedSolves and validates single-, multi-, and many-objective optimization with pymoo, including NSGA-II, NSGA-III, MOEA/D, constraints, Pareto approximations, reference directions, and ZDT/DTLZ benchmarks for engineering and research problems.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 13.8 KB |
| references/algorithms.md | markdown | 9.2 KB |
| references/constraints_mcdm.md | markdown | 9.0 KB |
| references/lifecycle.md | markdown | 5.2 KB |
| references/operators.md | markdown | 4.9 KB |
| references/parallelization.md | markdown | 2.6 KB |
| references/problems.md | markdown | 7.6 KB |
| references/quick_start_workflows.md | markdown | 7.3 KB |
| references/visualization.md | markdown | 4.7 KB |
| scripts/custom_problem_example.py | script | 5.0 KB |
| scripts/decision_making_example.py | script | 6.0 KB |
| scripts/many_objective_example.py | script | 2.3 KB |
| scripts/multi_objective_example.py | script | 1.7 KB |
| scripts/single_objective_example.py | script | 1.6 KB |
references/parallelization.md
View on GitHub# Parallel evaluation (pymoo 0.6.2)
Use a runner when individual evaluations dominate overhead. `ElementwiseProblem`
sets `elementwise=True`; no `elementwise_evaluation` argument is needed. Its
`_evaluate` receives one candidate, while the runner schedules multiple candidates.
Vectorized `Problem` evaluates a matrix directly and may be faster for NumPy models.
## Threads or processes
```python
from multiprocessing.pool import ThreadPool
from pymoo.algorithms.soo.nonconvex.ga import GA
from pymoo.core.problem import ElementwiseProblem
from pymoo.optimize import minimize
from pymoo.parallelization.starmap import StarmapParallelization
class MyProblem(ElementwiseProblem):
def __init__(self, **kwargs):
super().__init__(n_var=3, n_obj=1, xl=-5, xu=5, **kwargs)
def _evaluate(self, x, out, *args, **kwargs):
out["F"] = (x**2).sum()
if __name__ == "__main__":
with ThreadPool(2) as pool:
problem = MyProblem(elementwise_runner=StarmapParallelization(pool.starmap))
result = minimize(problem, GA(pop_size=20), ("n_gen", 5), seed=1)
```
The native test checks serial/thread equality on deterministic evaluations.
For CPU-bound Python functions, substitute `multiprocessing.get_context("spawn").Pool(2)`
inside the main guard. Keep worker classes/functions at importable module scope.
That process-pool variant is illustrative; it was not executed in this refresh.
Threads suit I/O and native code that releases the GIL. Avoid nested oversubscription
from BLAS or model runtimes. Context managers release workers on exceptions.
## Joblib
```python
from pymoo.parallelization.joblib import JoblibParallelization
runner = JoblibParallelization(n_jobs=2, backend="threading")
problem = MyProblem(elementwise_runner=runner)
result = minimize(problem, GA(pop_size=20), ("n_gen", 5), seed=1)
```
Install `joblib` separately. The constructor accepts `n_jobs` and joblib keyword
arguments; it does not take a lambda that manually wraps `Parallel`. The default
backend is joblib's process-based `loky`. Threading was executed here; distributed
and GPU evaluation require separate validation for the user's model and resources.
Optimizer seeds do not automatically seed stochastic external simulations or
worker-local RNGs. Supply deterministic per-evaluation seeds when appropriate and
record the evaluation protocol. Serialization of an algorithm does not preserve
live pools, remote clients, or external simulator state; recreate those resources
when resuming a checkpoint.
Sources: [Starmap](https://pymoo.org/parallelization/starmap.html),
[Joblib](https://pymoo.org/parallelization/joblib.html), and released 0.6.2 source;
reviewed 2026-10-01.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |
| fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks. | Stars: 115k | Updated | Claude CodeCodex | Verified | Install |