nexuml.execution.ray¶
Thin Ray Train execution for the canonical NexuML Lightning session.
Bases: RuntimeError
Raised when a Ray execution request cannot be represented cleanly.
train_loop_per_worker(config: dict[str, Any]) -> None
Run one Ray worker through the normal NexuML session lifecycle.
Raises:
| Type | Description |
|---|---|
RayExecutionError
|
If Ray Train is unavailable in the worker environment. |
run_ray(scenario: ScenarioSpec) -> Any
Execute one resolved scenario with Ray Train and return Ray's native Result.
Returns:
| Type | Description |
|---|---|
Any
|
The |
Raises:
| Type | Description |
|---|---|
RayExecutionError
|
If the scenario cannot be executed safely with Ray or Ray dependencies are unavailable. |