US 2020/0348912 A1 · OPTIMIZED MACHINE LEARNING PIPELINE EXECUTION BASED ON EXECUTION GRAPHS

Abstract

Embodiments provide a machine learning framework that enables developers to author and deploy machine learning pipelines into their applications regardless of the programming language in which the applications are structured. The framework may provide a programming language-specific API that enables the application to call a plurality of operators provided by the framework. The framework provides any number of APIs, each for a different programming language. The pipeline generated via the application is represented as an execution graph comprising node(s), where each node represents a particular operator. When a pipeline is submitted for execution, calls to the operators are detected, and nodes corresponding to the operators are generated for the execution graph. Once the execution graph is complete, the execution graph is provided to an execution engine, which reconstructs a corresponding execution graph that is structured in accordance with the operators' programming language and executes the reconstructed execution graph.

Claims20

Dependent claims

A method, comprising:

detecting calls to an application programming interface (API), each call being configured to invoke a particular operator in a plurality of operators and being structured in accordance with a first programming language, each of the operators in the plurality of operators being configured to perform a particular task with respect to a machine learning pipeline of a machine learning model;
for each detected call, generating a corresponding node that corresponds to the operator invoked thereby in a first execution graph that is representative of the machine learning pipeline and is structured in accordance with the first programming language;
providing the first execution graph to a graph execution engine;
generating, by the graph execution engine, a second execution graph that is structured in accordance with a second programming language;
and
executing, by the graph execution engine, the second execution graph.

3 more elements

A system, comprising:

at least one processor circuit;
and
at least one memory that stores program code configured to be executed by the at least one processor circuit, the program code comprising:
an API detector configured to detect calls to an application programming interface (API), each call being configured to invoke a particular operator in a plurality of operators and being structured in accordance with a first programming language, each of the operators in the plurality of operators being configured to perform a particular task with respect to a machine learning pipeline of a machine learning model;
a graph builder configured to, for each detected call, generate a corresponding node that corresponds to the operator invoked thereby in a first execution graph that is representative of the machine learning pipeline and is structured in accordance with the first programming language;
and
a graph execution engine configured to:
receive the first execution graph to a graph execution engine;
generate a second execution graph that is structured in accordance with a second programming language;
and
execute the second execution graph.

8 more elements

A computer-readable storage medium having program instructions recorded thereon that, when executed by at least one processor of a computing device, perform a method, the method comprising:

detecting calls to an application programming interface (API), each call being configured to invoke a particular operator in a plurality of operators and being structured in accordance with a first programming language, each of the operators in the plurality of operators being configured to perform a particular task with respect to a machine learning pipeline of a machine learning model;
for each detected call, generating a corresponding node that corresponds to the operator invoked thereby in a first execution graph that is representative of the machine learning pipeline and is structured in accordance with the first programming language;
providing the first execution graph to a graph execution engine;
generating, by the graph execution engine, a second execution graph that is structured in accordance with a second programming language;
and
executing, by the graph execution engine, the second execution graph.

3 more elements

Dependencies are inferred from the claim text.

Optimized machine learning pipeline execution based on execution graphs

Life of this patent

Optimized machine learning pipeline execution based on execution graphs

Life of this patent

Abstract

Claims20