run_graph_with_stream_async
Function Usage
Runs a graph asynchronously.
Prototype
1 | run_graph_with_stream_async(graph_id: int, stream: int, inputs: List[Tensor]) -> List[Tensor] |
Parameters
Parameter |
Input/Output |
Description |
|---|---|---|
graph_id |
Input |
Graph ID |
stream |
Input |
Stream on which the graph is run |
inputs |
Input |
Input tensor list, which is the memory space on the device |
Returns
(List[Tensor]) Output tensor list, which is the memory space on the device
Restrictions
- If graph_id is not an integer, TypeError is thrown.
- If stream is not an integer, TypeError is thrown.
- If any element in inputs is not of the tensor type, TypeError is thrown.
- If the execution fails, RuntimeError is thrown.
- Before calling this API, you need to create a stream by calling acl.rt.create_stream() provided by ACL. Only the scenario where the stream uses the default context is supported. Before obtaining the output result, you need to call acl.rt.synchronize_stream to ensure that the tasks on the stream have been executed. For details about the API, see Stream Management.
Parent topic: Session APIs