Decorators API Reference

@experiment_manager

The main decorator for managing experiment execution.

Syntax

@experiment_manager(
    times: int = 1,
    parallel: bool = False,
    merge_config: Optional[Dict] = None,
    experiment_name: Optional[str] = None
)

Parameters

• times: Number of experiment runs
• parallel: Enable parallel execution
• merge_config: Configuration for merging results
• experiment_name: Optional custom experiment name

Example

from orruns.decorators import experiment_manager

@experiment_manager(
    times=5,
    parallel=True,
    merge_config={
        "scalars": ["objective", "time"],
        "distributions": ["gap"]
    }
)
def optimization_experiment(tracker):
    # Your experiment code
    params = {
        "method": "barrier",
        "threads": 4
    }
    tracker.log_params(params)

    result = solve_problem(params)
    return {
        "objective": result.objective,
        "time": result.time,
        "gap": result.gap
    }

Result Merging

The merge_config parameter supports: - scalars: Metrics to aggregate (mean, std, min, max) - distributions: Metrics to analyze as distributions - artifacts: Artifact handling configuration

Common Use Cases

1. Simple Repetition

@experiment_manager(times=3)
def basic_experiment(tracker):
    # Run 3 times sequentially
    pass

2. Parallel Execution

@experiment_manager(
    times=10,
    parallel=True
)
def parallel_experiment(tracker):
    # Run 10 times in parallel
    pass

3. Result Analysis

@experiment_manager(
    times=5,
    merge_config={
        "scalars": ["objective"],
        "distributions": ["iterations"]
    }
)
def analysis_experiment(tracker):
    # Results will be automatically analyzed
    pass