Quiz dossier / Reinforcement Learning

Distributed RL & Ray Deep Dive

Advanced drill on scaling reinforcement learning: Ray's task/actor/object-store model, RLlib abstractions, actor-learner architectures (A3C, Ape-X, IMPALA, SEED RL, R2D2), synchronous vs asynchronous trade-offs, policy lag, and deployment on Kubernetes.

22 questionsEstimated 45 min0 flagged

Session status

Draft

Answered: 0/22
Score: 0/22

Question 1

1 point(s)

Match each Ray primitive to its definition

Ray's core model underpins RLlib.

A stateless remote function invocation scheduled across the cluster

A stateful remote worker (a remote class instance) that keeps state between calls

Shared-memory store of immutable objects, enabling zero-copy reads on a node

Cluster-wide metadata service tracking actors, objects and resources

The top-level program that launches the computation