Pipeline processing system and pipeline processing program

Abstract

To provide a pipeline processing system with improved maintainability.SOLUTION: A pipeline processing system which processes a pipeline connecting multiple blocks that are executable programs with ability to load a trained neural network model for inference processing comprises: storage means configured to store each block file configuring the pipeline; storage means configured to store pipeline configuration information which defines a configuration of the pipeline and includes information about all blocks constituting the pipeline and information about the order of executing the blocks constituting the pipeline; and pipeline processing control means configured to load and execute each of the blocks included in the pipeline using data input from programs outside the pipeline according to the pipeline configuration information, and return results of executing the pipeline processing to the external programs.SELECTED DRAWING: Figure 3

Claims

1. A pipeline processing system that performs pipeline processing by connecting multiple blocks, which are executable programs that have the function of loading a pre-trained neural network model for inference processing, Block file storage means for storing the files of each block constituting the pipeline, Configuration information storage means for storing pipeline configuration information, which includes information defining the configuration of the pipeline, including information about all blocks constituting the pipeline and information about the execution order of each block constituting the pipeline. A pipeline processing system comprising: a pipeline processing control means that loads and executes each of the blocks included in the pipeline using data input from an external program (hereinafter referred to as "external program") to the pipeline according to the pipeline configuration information, and returns the execution result of the pipeline processing to the external program.

2. The aforementioned pipeline processing system is equipped with multiple information terminals, The pipeline processing system according to claim 1, characterized in that each of the information terminals is equipped with a transmission and reception means capable of sending the output of each instance of a block executed on its own machine to another information terminal that uses it as input via a network, and receiving the output from each instance of a block in another information terminal as input to each instance of a block on its own machine via a network.

3. The aforementioned pipeline processing system further comprises a server, The information terminal that executes the external program includes terminal-side configuration information storage means for storing terminal-side pipeline configuration information, which includes at least information about blocks to be executed on the information terminal and information about the execution order of these blocks from the pipeline configuration information. The pipeline processing system according to claim 2, further comprising a user interface means for editing the pipeline configuration information and writing the terminal-side pipeline configuration information from the pipeline configuration information to the terminal-side configuration information storage means of the information terminal.

4. The aforementioned pipeline processing system further comprises a server, The pipeline processing system according to claim 2 or 3, characterized in that the server includes performance display means for receiving and displaying performance information of the processing of each instance of the block from each of the information terminals.

5. The aforementioned pipeline processing system further comprises a server, The information terminal that executes the external program includes terminal-side configuration information storage means for storing terminal-side pipeline configuration information, which includes at least information about blocks to be executed on the information terminal and information about the execution order of these blocks from the pipeline configuration information. The pipeline processing system according to claim 2, characterized in that the server receives performance information of the processing of each block instance from each of the information terminals, automatically edits the pipeline configuration information based on this performance information, and writes the terminal-side pipeline configuration information from this pipeline configuration information to the terminal-side configuration information storage means of the information terminal.

6. The pipeline processing system according to claim 1, further comprising a code library storage means incorporated into the external program and storing a library of source code necessary for executing blocks included in the pipeline in the execution environment of the external program.

7. A pipeline processing system that performs pipeline processing by combining multiple blocks, The aforementioned multiple blocks are computer-executable electronic files that perform data input and output, The aforementioned pipeline processing system consists of a server and a device, The aforementioned server, A display means for displaying the pipeline configuration by arranging icons representing the aforementioned multiple blocks on the screen, A modification means that modifies the configuration of the pipeline displayed on the display means based on user input, A creation means that creates a configuration file showing the configuration of the pipeline based on the configuration of the pipeline displayed on the display means, A selection means for selecting the device to which the configuration file will be deployed, The system includes a deployment means for deploying the configuration file to the device selected by the selection means, The configuration file is a file of information defining the configuration of the pipeline, and includes information about the blocks that constitute the pipeline and information about the execution order of each block that constitutes the pipeline. The device described above, When the configuration file is deployed by the deployment means, the acquisition means obtains multiple blocks that constitute the pipeline indicated therein. A pipeline processing system comprising execution means that sequentially executes the acquired plurality of blocks according to the configuration file.

8. A pipeline processing program that performs pipeline processing by connecting multiple blocks, which are executable programs that have the function of loading a pre-trained neural network model for inference processing, Information processing equipment, Block file storage means for storing the files of each block constituting the pipeline, Configuration information storage means for storing pipeline configuration information, which includes information defining the configuration of the pipeline, including information about all blocks constituting the pipeline and information about the execution order of each block constituting the pipeline. A pipeline processing program that functions as a pipeline processing control means, loading and executing each of the blocks included in the pipeline using data input from an external program (hereinafter referred to as "external program") in accordance with the pipeline configuration information, and returning the execution result of the pipeline processing to the external program.

9. The pipeline processing program according to claim 8, further characterized in that the information processing device is incorporated into the external program and functions as a code library storage means for storing a library of source code necessary for executing blocks included in the pipeline in the execution environment of the external program.

Images