public SimpleBayesianOptimizerProxy(ModelsDatabase modelsDatabase, OptimizationProblem optimizationProblem, Guid?optimizerId)
{
this.modelsDatabase = modelsDatabase;
this.optimizationProblem = optimizationProblem;
OptimizerId = optimizerId;
optimizerExecutionContext = new SimpleBayesianOptimizerExecutionContext()
{
OptimizerId = optimizerId,
};
optimizerExecutionContext.ModelVersions.Add(0);
}
public static void Main(string[] args)
{
// TODO: use some proper arg parser. For now let's keep it simple.
//
string executableFilePath = null;
string modelsDatabaseConnectionDetailsFile = null;
foreach (string arg in args)
{
if (Path.GetExtension(arg) == ".exe")
{
executableFilePath = arg;
}
else if (Path.GetExtension(arg) == ".json")
{
modelsDatabaseConnectionDetailsFile = arg;
}
}
Console.WriteLine("Mlos.Agent.Server");
TargetProcessManager targetProcessManager = null;
// Since the models database and optimizer factory are part of the "intelligence" it belongs to the Mlos.Agent.Server.
//
ModelsDatabase modelsDatabase = new ModelsDatabase(modelsDatabaseConnectionDetailsFile);
SimpleBayesianOptimizerFactory optimizerFactory = new SimpleBayesianOptimizerFactory(modelsDatabase);
// Since component assemblies need to use the optimizerFactory (at least for now) we put it in the GlobalProperties.
//
MlosContext.OptimizerFactory = optimizerFactory;
// Create circular buffer shared memory before running the target process.
//
MainAgent.InitializeSharedChannel();
// Active learning, almost done. In active learning the MlosAgentServer controls the workload against the target component.
//
if (executableFilePath != null)
{
targetProcessManager = new TargetProcessManager(executableFilePath: executableFilePath);
targetProcessManager.StartTargetProcess();
}
var cancelationTokenSource = new CancellationTokenSource();
Task grpcServerTask = CreateHostBuilder(Array.Empty <string>()).Build().RunAsync(cancelationTokenSource.Token);
Console.WriteLine("Starting Mlos.Agent");
Task mlosAgentTask = Task.Factory.StartNew(
() => MainAgent.RunAgent(),
TaskCreationOptions.LongRunning);
if (targetProcessManager != null)
{
targetProcessManager.WaitForTargetProcessToExit();
targetProcessManager.Dispose();
MainAgent.UninitializeSharedChannel();
}
Console.WriteLine("Waiting for Mlos.Agent to exit");
mlosAgentTask.Wait();
mlosAgentTask.Dispose();
cancelationTokenSource.Cancel();
grpcServerTask.Wait();
grpcServerTask.Dispose();
cancelationTokenSource.Dispose();
Console.WriteLine("Mlos.Agent exited.");
}
public SimpleBayesianOptimizerFactory(ModelsDatabase modelsDatabase)
{
this.modelsDatabase = modelsDatabase;
}