ray.rllib.utils.exploration.random.Random.postprocess_trajectory

Random.postprocess_trajectory(policy: Policy, sample_batch: SampleBatch, tf_sess: tf.Session | None = None)

Handles post-processing of done episode trajectories.

Changes the given batch in place. This callback is invoked by the sampler after policy.postprocess_trajectory() is called.

Parameters:

• policy – The owning policy object.

• sample_batch – The SampleBatch object to post-process.

• tf_sess – An optional tf.Session object.