Granted Patent B2

US 12,481,672 · App. 18/678,996 · Granted Nov 25, 2025

Inventors: Antariksha Bhaduri (Karnataka, IN); Kripa Kanchana Sivakumar (Seattle, WA)

Assignee: Oracle International Corporation

G06F16/258

Abstract

Claims (90)

1 . A computer-implemented method comprising:

receiving, by a metrics management system and from a first machine-learning framework, first metric data for a first metric related to machine-learning processing;

determining, by the metrics management system, a first metric data type for the first metric data, wherein the first metric data type is determined from a set of different metric data types;

identifying, by the metrics management system and based upon the first metric data type and a source of the first metric data, a first adapter for processing the received first metric data, wherein the first adapter is configured to convert data in the first metric data type to metric data in a common metrics format specified by a common metrics schema;

generating, by the metrics management system and by using the first adapter, the metric data in the common metrics format for the first metric based upon the first metric data, the generating comprising:

identifying a first type converter;

converting the first metric data to the metric data in the common metrics format using the first type converter; and

storing, by the metrics management system, the metric data in the common metrics format for the first metric in a memory store;

receiving, from a first requestor, a first request for the first metric;

determining, based upon the first request, that data is to be provided in a first requestor format, wherein the first requestor format is different from the first metric data type and from the common metrics format; and

responsive to determining that the data is to be provided in the first requestor format;

determining, by the metrics management system, a second type converter configured to convert the first metric from the common metrics format to the first requestor format;

converting, by the metrics management system and by using the second type converter, the metric data in the common metrics format for the first metric from the common metrics format to the first requestor format; and

providing, by the metrics management system, the first metric to the first requestor in the first requestor format using the stored first common metrics format data.

2 . The computer-implemented method of claim 1 , wherein the common metrics format comprises a plurality of fields, wherein the plurality of fields comprises a metric description, a variable count, a variable dimension, a variable name, a variable type, a unit, a variable data type, raw data, and a type converter, wherein the variable type is selected from a group consisting of nominal, categorical, and continuous, wherein the variable data type is selected from a group consisting of integer, float, fraction, percentage, array, string, and tensor, wherein the raw data is a blob representation of the received first metric data, wherein the computer-implemented method further comprises identifying a plurality of data included in the first metric data, wherein the plurality of data corresponds to the plurality of fields, and wherein the method further comprises identifying, for each field of the plurality of fields, corresponding data included in the plurality of data to convert the first metric data.

3 . The computer-implemented method of claim 1 , further comprising:

receiving second metric data for a second metric from a second machine-learning framework, the second metric data having a second metric data type; and

updating the set of different metric data types to add the second metric data type in response to determining that a second metric data type of the second metric data is different from existing metric data types included in the set of different metric data types.

4 . The computer-implemented method of claim 1 , wherein identifying the first adapter comprises identifying the first adapter based on the source of the first metric data, the method further comprising determining the source of the first metric data.

5 . The computer-implemented method of claim 1 , further comprising:

receiving second metric data for a second metric related to machine-learning processing;

determining a second metric data type for the second metric;

identifying a second adapter for processing the received second metric;

generating second common metrics format data for the second metric based upon the second metric data and using the second adapter, the generating comprising converting the second metric data from the second metric data type to the common metrics format;

storing the second common metrics format data in the memory store; and

responsive to a second request from a second requestor requesting the second metric in a second requestor format, providing the second metric to the second requestor in the second requestor format using the stored second common metrics format data.

6 . The computer-implemented method of claim 5 , wherein receiving the second metric data comprises receiving the second metric data from a second particular source that is different than a first particular source associated with the first metric data, and wherein the second adapter is different than the first adapter.

7 . The computer-implemented method of claim 1 , wherein generating first common metrics format data comprises:

extracting predefined data from the first metric data for at least a subset of a plurality of fields included in the common metrics format;

using the first adapter to call a particular type converter to convert the first metric data from the first metric data type to the common metrics format; and

converting, using the particular type converter, the first metric data from the first metric data type into a blob or a list of list of strings for at least one field of the plurality of fields to prevent data from being lost during processing.

8 . A system comprising:

one or more processors; and

a memory coupled to the one or more processors, the memory storing a plurality of instructions executable by the one or more processors, the plurality of instructions comprising instructions executable by the one or more processors to cause the one or more processors to perform operations comprising:

receiving, from a first machine-learning framework, first metric data for a first metric related to machine-learning processing;

determining a first metric data type for the first metric data, wherein the first metric data type is determined from a set of different metric data types;

identifying, based upon the first metric data type and a source of the first metric data, a first adapter for processing the received first metric data, wherein the first adapter is configured to convert data in the first metric data type to metric data in a common metrics format specified by a common metrics schema;

generating, by using the first adapter, the metric data in the common metrics format for the first metric based upon the first metric data, the generating comprising:

identifying a first type converter; and

converting the first metric data to the metric data in the common metrics format using the first type converter;

the metric data in the common metrics format for the first metric in a memory store;

receiving, from a first requestor, a first request for the first metric

determining, based upon the first request, that data is to be provided in a first requestor format, wherein the first requestor format is different from the first metric data type and from the common metrics format; and

responsive to determining that the data is to be provided in the first requestor format;

determining a second type converter configured to convert the first metric from the common metrics format to the first requestor format;

converting, by using the second type converter, the metric data in the common metrics format for the first metric from the common metrics format to the first requestor format; and

providing the first metric to the first requestor in the first requestor format.

9 . The system of claim 8 , wherein the common metrics format comprises a plurality of fields, wherein the plurality of fields comprises a metric description, a variable count, a variable dimension, a variable name, a variable type, a unit, a variable data type, raw data, and a type converter, and wherein the variable type is selected from a group consisting of nominal, categorical, and continuous, wherein the variable data type is selected from a group consisting of integer, float, fraction, percentage, array, string, and tensor, and wherein the raw data is a blob representation of the received first metric data.

10 . The system of claim 9 , wherein the operations further comprise identifying a plurality of data included in the first metric data, wherein the plurality of data corresponds to the plurality of fields, and wherein the operations further comprise identifying, for each field of the plurality of fields, corresponding data included in the plurality of data to convert the first metric data.

11 . The system of claim 9 , wherein the operation of identifying the first adapter comprises identifying based on the source of the first metric data, the operations further comprising determining the source of the first metric data.

12 . The system of claim 9 , wherein the operations further comprise:

receiving second metric data for a second metric related to machine-learning processing;

determining a second metric data type for the second metric;

identifying a second adapter for processing the received second metric;

generating second common metrics format data for the second metric based upon the second metric data and using the second adapter, the generating comprising converting the second metric data from the second metric data type to the common metrics format;

storing the second common metrics format data in the memory store; and

responsive to a second request from a second requestor requesting the second metric in a second requestor format, providing the second metric to the second requestor in the second requestor format using the stored second common metrics format data.

13 . The system of claim 12 , wherein the operation of receiving the second metric data comprises receiving the second metric data from a second particular source that is different than a first particular source associated with the first metric data, and wherein the second adapter is different than the first adapter.

14 . The system of claim 8 , wherein the operation of generating first common metrics format data comprises:

extracting predefined data from the first metric data for at least a subset of a plurality of fields included in the common metrics format;

using the first adapter to call a particular type converter to convert the first metric data from the first metric data type to the common metrics format; and

converting, using the particular type converter, the first metric data from the first metric data type into a blob or a list of list of strings for at least one field of the plurality of fields to prevent data from being lost during processing.

15 . A non-transitory computer-readable memory storing a plurality of instructions executable by one or more processors, the plurality of instructions comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, comprising:

receiving, from a first machine-learning framework, first metric data for a first metric related to machine-learning processing;

determining a first metric data type for the first metric data, wherein the first metric data type is determined from a set of different metric data types;

identifying, based upon the first metric data type and a source of the first metric data, a first adapter for processing the received first metric data, wherein the first adapter is configured to convert data in the first metric data type to metric data in a common metrics format specified by a common metrics schema;

generating, by using the first adapter, the metric data in the common metrics format for the first metric based upon the first metric data, the generating comprising:

identifying a first type converter; and

converting the first metric data to the metric data in the common metrics format using the first type converter;

storing the metric data in the common metrics format for the first metric in a memory store;

receiving, from a first requestor, a first request for the first metric;

determining, based upon the first request, that data is to be provided in a first requestor format, wherein the first requestor format is different from the first metric data type and from the common metrics format; and

responsive to determining that the data is to be provided in the first requestor format;

determining a second type converter configured to convert the first metric from the common metrics format to the first requestor format;

converting, by using the second type converter, the metric data in the common metrics format for the first metric from the common metrics format to the first requestor format; and

providing, the first metric to the first requestor in the first requestor format.

16 . The non-transitory computer-readable memory of claim 15 , wherein the common metrics format comprises a plurality of fields, wherein the plurality of fields comprises a metric description, a variable count, a variable dimension, a variable name, a variable type, a unit, a variable data type, raw data, and a type converter, and wherein the variable type is selected from a group consisting of nominal, categorical, and continuous, wherein the variable data type is selected from a group consisting of integer, float, fraction, percentage, array, string, and tensor, and wherein the raw data is a blob representation of the received first metric data.

17 . The non-transitory computer-readable memory of claim 16 , wherein the operations further comprise identifying a plurality of data included in the first metric data, wherein the plurality of data corresponds to the plurality of fields, and wherein the operation further comprise identifying, for each field of the plurality of fields, corresponding data included in the plurality of data to convert the first metric data.

18 . The non-transitory computer-readable memory of claim 16 , wherein the operation of, identifying the first adapter comprises identifying based on the source of the first metric data, the operations further comprising determining the source of the first metric data.

19 . The non-transitory computer-readable memory of claim 16 , wherein the operations further comprise:

receiving second metric data for a second metric related to machine-learning processing;

determining a second metric data type for the second metric;

identifying a second adapter for processing the received second metric;

generating second common metrics format data for the second metric based upon the second metric data and using the second adapter, the generating comprising converting the second metric data from the second metric data type to the common metrics format;

storing the second common metrics format data in the memory store; and

responsive to a second request from a second requestor requesting the second metric in a second requestor format, providing the second metric to the second requestor in the second requestor format using the stored second common metrics format data, wherein the operation of receiving the second metric data comprises receiving the second metric data from a second particular source that is different than a first particular source associated with the first metric data, and wherein the second adapter is different than the first adapter.

20 . The non-transitory computer-readable memory of claim 15 , wherein the operation of generating first common metrics format data comprises:

extracting predefined data from the first metric data for at least a subset of a plurality of fields included in the common metrics format;

using the first adapter to call a particular type converter to convert the first metric data from the first metric data type to the common metrics format; and

converting, using the particular type converter, the first metric data from the first metric data type into a blob or a list of list of strings for at least one field of the plurality of fields to prevent data from being lost during processing.

Assignments (1)

Continuity (2)

References Cited (42)