td3_pendulum

This is an example to train a task with TD3 algorithm.

Here, we create a gym environment InvertedDoublePendulum and use a TD3 with 1M steps.

Results:

AverageReturn: 250 RiseTime: epoch 499

td3_pendulum(ctxt=None, seed=1)

Wrap TD3 training task in the run_task function.

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.