Hint
Example code: https://github.com/cselab/korali/tree/master/examples/features/multiple.experiments/
Running Multiple Experiments¶
In this tutorial we show how you can execute a series of experiments, in order to benefit from Korali’s oversubscription capabilities.
Create a series of experiments:
for i in range(8):
e = korali.Experiment()
e["Problem"]["Type"] = "Evaluation/Bayesian/Inference/Reference"
e["Problem"]["Likelihood Model"] = "Additive Normal"
e["Problem"]["Reference Data"] = getReferenceData()
e["Problem"]["Computational Model"] = lambda sampleData: model(sampleData, getReferencePoints())
# Configuring CMA-ES parameters
e["Solver"]["Type"] = "Optimizer/CMAES"
...
Set Experiment Vector¶
We can store experiments in a list eList:
# Adding Experiment to vector
eList.append(e)
Run Experiment Vector¶
We can runn all experiments in one Korali application
# Running first 100 generations
k.run(eList)
Running¶
We are now ready to run our example: python3 ./run-cmaes
Resuming Previous Run¶
Runs with multiple Korali experiments can also be resumed from a previous execution. For more information, see Checkpoint / Resume.