Huawei uses machine translation combined with human proofreading to translate this document to different languages in order to help you better understand the content of this document.
Note: Even the most advanced machine translation cannot match the quality of professional translators.
Huawei shall not bear any responsibility for translation accuracy and it is recommended that you refer to the English document (a link for which has been provided).
Scenarios with Callback
Figure 4-16 Synchronization flowchart for model inference with callback
![]()
Figure 4-17 Synchronization flowchart for model inference and operator calling with callback
![]()
The callback flow is briefly described as follows:
- The user needs to create a callback function in advance to obtain and process the model inference or operator execution result.
- The user needs to create a thread and customize a thread function in advance. Once acl.rt.process_report is called in the thread function, the created callback function will be triggered after a specified period of time.
- Call acl.rt.subscribe_report to specify the thread for processing the callback function in the stream created in advance.
- Call acl.rt.launch_callback to add the created callback function to be executed on the host or device to the task stream.
- Call acl.rt.unsubscribe_report to unsubscribe a thread. (The callback function in the stream is no longer processed by the specified thread.)
- In the asynchronous inference scenario with callback processing, acl.rt.synchronize_stream needs to be called before the stream is destroyed to ensure all tasks in a stream are complete and all inference results are handled by the callback function.