Inference Task Lifecycle State Management System

Abstract

This case discloses a lifecycle state management system for inference tasks, comprising an inference task setting module, an inference task management module, an inference execution module, at least one device monitoring module, multiple state definition modules, a state transition control module, a real-time state recording module, and a user interface module. The state definition modules define multiple state nodes for each module's lifecycle based on the real-time inference task. The state transition control module determines one of the corresponding state nodes as a current state node within the lifecycle based on the current log file of the real-time inference task, and determines whether the real-time inference task needs a state transition based on the updated current log file. The real-time state recording module is signal-connected to the state transition control module to record the state transitions of all state nodes. The user interface module provides a user interface to display the real-time inference task and its corresponding current state node.

Claims

1. A lifecycle state management system for an inference task, comprising: an inference task setting module for setting and creating an instant inference task; and an inference task management module for scheduling the instant inference task; An inference execution module for executing the real-time inference task; at least one device monitoring module for monitoring the device executing the real-time inference task; multiple state definition modules, signal-connected to the inference task setting module, the inference task management module, the inference execution module, and the at least one device monitoring module, respectively, to define multiple state nodes for a lifecycle according to the real-time inference task; a state transition control module, signal-connected to the multiple state definition modules, to determine one of the multiple state nodes as a current state node in the lifecycle according to a current log file of the real-time inference task, and to determine whether the real-time inference task should undergo a state transition according to the updated current log file; and a real-time state recording module, signal-connected to the state transition control module, to record the current state node, each state transition, and the current state node updated by the real-time inference task after the state transition. And a user interface module, which is signal-connected to the state transition control module, to provide a user interface for displaying the real-time inference task and its corresponding current state node.

2. The lifecycle state management system for the inference task as described in claim 1 further includes an anomaly detection and reporting module, which is signal-connected to the state transition control module to provide real-time reporting when the real-time inference task experiences an anomaly or stalls, and to add a prominent icon to the corresponding current state node on the user interface through the state transition control module.

3. The lifecycle state management system for the inference task as described in claim 1, wherein the at least one device monitoring module further includes a camera monitoring module and a hardware and network monitoring module.

4. A lifecycle state management system for an inference task as described in claim 1, wherein the plurality of state nodes is an idle node, a scheduled node, a deployed node, a starting node, an initialization node, an inference execution node, and a completed node.

5. The lifecycle state management system for the inference task as described in claim 4, wherein the state transition control module takes the idle node as the initial current state node, and when the state transition control module determines that the real-time inference task needs to perform the state transition based on the updated current log file, the plurality of state nodes are sequentially used as the updated current state node according to the order of the scheduled node, the deployed node, the starting node, the initialization node, the inference execution node, and the completed node.

6. The lifecycle state management system for the inference task as described in claim 5, wherein when the state transition control module changes the state of the real-time inference task from an idle state to a scheduled state according to the updated current log file, the scheduled node corresponding to the scheduled state is presented in a recognizable visual manner on the user interface through the user interface module.

7. The lifecycle state management system for inference tasks as described in claim 5, wherein when the state transition control module changes the state of the real-time inference task from a scheduled state to a deployed state according to the updated current log file, the deployed node corresponding to the deployed state is presented in a recognizable visual manner on the user interface through the user interface module.

8. The lifecycle state management system for the inference task as described in claim 5, wherein when the state transition control module changes the state of the real-time inference task from a deployed state to a starting state according to the updated current log file, the starting node corresponding to the starting state is presented in a recognizable visual manner on the user interface through the user interface module.

9. The lifecycle state management system for the inference task as described in claim 5, wherein when the state transition control module changes the state of the real-time inference task from a starting state to an initialization state according to the updated current log file, the initialization node corresponding to the initialization state is presented in a recognizable visual manner on the user interface through the user interface module.

10. The lifecycle state management system for inference tasks as described in claim 5, wherein when the state transition control module changes the state of the real-time inference task from an initial state to an inference execution state according to the updated current log file, the inference execution node corresponding to the inference execution state is presented in a recognizable visual manner on the user interface through the user interface module.

11. The lifecycle state management system for the inference task as described in claim 5, wherein when the state transition control module changes the state of the real-time inference task from an inference execution state to a completed state according to the updated current log file, the completion node corresponding to the completed state is presented in a recognizable visual manner on the user interface through the user interface module.

12. A lifecycle state management system for an inference task as described in claim 5, wherein when the node in the inference process is the current state node, the current state node is updated to the deployed node according to a redeployment operation.

13. A lifecycle state management system for an inference task as described in claim 5, wherein when the node in the inference process is the current state node, the current state node is updated to the initialization node according to a reinitialization operation.

14. The inference task lifecycle state management system as described in claim 1, wherein the inference task setting module, the inference task management module, the inference operation module, the at least one device monitoring module, the multiple state definition module, the state transition control module, the real-time state recording module, and the user interface module are located in an electronic device and are executed through the electronic device.

15. A lifecycle state management system for inference tasks as described in claim 14, wherein the electronic device is a cloud server or a local computer.

Images