static void RunRepeat()
{
RepeatObservationEnvironment.SanityCheck();
var random = new Random();
const int RepeatAgents = 3;
ndarray RepeatRandomActionSampler()
=> ndarray.FromList(Range(0, RepeatAgents)
.Select(_ => (float)random.NextDouble() * 2 - 1)
.ToList())
.reshape(new int[] { RepeatAgents, 1 })
.AsArray <float>();
SoftActorCritic.SoftActorCritic.Run(new RepeatObservationEnvironment(RepeatAgents),
agentGroup: null,
actorCriticFactory: ActorCriticFactory,
observationDimensions: 1,
actionDimensions: 1,
actionLimit: 1,
feedFrames: 1,
hiddenSizes: new int[] { 32 },
maxEpisodeLength: 256,
replaySize: 1024 * 1024 / 8,
learningRate: 2e-4f,
startSteps: 100,
actionSampler: RepeatRandomActionSampler);
}
public static void SanityCheck()
{
// sanity check
var env = new RepeatObservationEnvironment(agents: 3);
env.Reset();
env.Step();
for (int episode = 0; episode < 100; episode++)
{
var observation = (ndarray)env.GetStepResult(null).Item1.obs[0];
env.SetActions(null, observation);
env.Step();
var step = env.GetStepResult(null);
var success = step.Item1.reward >= 1.99f;
bool allPass = success.all();
Trace.Assert(allPass);
}
}
