ray.rllib.utils.exploration.exploration.Exploration.before_compute_actions#

Exploration.before_compute_actions(*, timestep: numpy.array | jnp.ndarray | tf.Tensor | torch.Tensor | int | None = None, explore: numpy.array | jnp.ndarray | tf.Tensor | torch.Tensor | bool | None = None, tf_sess: tf.Session | None = None, **kwargs)[source]#

Hook for preparations before policy.compute_actions() is called.

Parameters:

  • timestep – An optional timestep tensor.

  • explore – An optional explore boolean flag.

  • tf_sess – The tf-session object to use.

  • **kwargs – Forward compatibility kwargs.