nv_ingest.framework.orchestration.ray.util.system_tools package#
Submodules#
nv_ingest.framework.orchestration.ray.util.system_tools.memory module#
- nv_ingest.framework.orchestration.ray.util.system_tools.memory.estimate_actor_memory_overhead(
- actor_class,
- iterations=1,
- stabilization_threshold=1048576,
- wait_time=2,
- actor_args=None,
- actor_kwargs=None,
Estimate the additional system memory overhead when launching a Ray actor of the given actor_class.
- Parameters:
actor_class – A Ray remote actor class.
iterations (int) – Number of measurement iterations.
stabilization_threshold (int) – Maximum difference (in bytes) between min and max measurements to
stable. (consider results)
wait_time (float) – Seconds to wait after spawning or killing an actor for memory to stabilize.
actor_args (list) – Positional arguments to pass to the actor’s remote() call.
actor_kwargs (dict) – Keyword arguments to pass to the actor’s remote() call.
- Returns:
Estimated average overhead in bytes for replicating the actor.
- Return type:
float