Hello community. I was wondering if I could get some help here. I am trying to send tasks to raytaskrunner and I understand I can do as follows:

from prefect import flow, task
from prefect_ray.task_runners import RayTaskRunner

@task
def parallel_task(task_id: int):
    return f"Result from task {task_id}"

@task
def join_task(results):
    combined_result = " | ".join(results)
    return f"Joined result: {combined_result}"

@flow(task_runner=RayTaskRunner(address="<ray://raycluster-kuberay-head-svc.kuberay.svc.cluster.local:10001>"))
def fork_and_join_workflow():
    # Fork: Execute 50 parallel tasks
    parallel_results = [parallel_task.submit(i) for i in range(50)]
    
    # Join: Aggregate results from all tasks
    results = [result.result() for result in parallel_results]
    joined_result = join_task(results)
    
    print(joined_result)

if __name__ == "__main__":
    fork_and_join_workflow()

Now I want to allocate 1 GPU and 4 CPU for each task that is submitted over to ray, how can I make sure of that?! In regular ray, I can just decorate a function with @ray.remote(num_gpus=1, num_cpus=4), but I cannot do that with prefect tasks?