cma_es_cartpole

This is an example to train a task with CMA-ES.

Here it runs CartPole-v1 environment with 100 epoches.

Results:

AverageReturn: 100 RiseTime: epoch 38 (itr 760),

but regression is observed in the course of training.

cma_es_cartpole(ctxt=None, seed=1)

Train CMA_ES with Cartpole-v1 environment.

Parameters

  • ctxt (garage.experiment.ExperimentContext) – The experiment configuration used by Trainer to create the snapshotter.

  • seed (int) – Used to seed the random number generator to produce determinism.