ddpg_pendulum

This is an example to train a task with DDPG algorithm.

Here it creates a gym environment InvertedDoublePendulum. And uses a DDPG with 1M steps.

Results:

AverageReturn: 250 RiseTime: epoch 499

ddpg_pendulum(ctxt=None, seed=1)

Train DDPG with InvertedDoublePendulum-v2 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.