Do Training and evaluation on GPU

RLlib
1

How severe does this issue affect your experience of using Ray?

  • High: It blocks me to complete my task.

I am trying to train an RL agent using Rllib. Specifically I want the training and evaluation to be done on GPU. The training runs fine on GPU but I got no GPU available for the evaluation. Here is how I specified the resource allocation and the evaluation.

.rollouts(num_rollout_workers=1, num_envs_per_worker=1, rollout_fragment_length=128)
 .resources(num_cpus_per_worker=1, num_gpus_per_worker=0.5/1)
    .evaluation(evaluation_interval=1, evaluation_duration=10, evaluation_duration_unit="episodes")
)