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

Here it runs CubeCrash-v0 environment with 100 iterations.

trpo_cubecrash(ctxt=None, seed=1, max_episode_length=5, batch_size=4000)

Train TRPO with CubeCrash-v0 environment.

  • 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.

  • max_episode_length (int) – Maximum length of a single episode.

  • batch_size (int) – Number of timesteps to use in each training step.