IP Library › Granted Patent US 12670382
Granted Patent B2
US 12670382 · App. 17/208,707 · Granted Jun 30, 2026

Monitoring operator compatibility within a deep learning framework

Inventors: Yamei Li (Beijing, CN); Xiang Lan (Beijing, CN); Tao Luo (Beijing, CN); Huihuang Zheng (Beijing, CN)
Assignee: Beijing Baidu Netcom Science and Technology Co., Ltd.
G06N3/08
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12670382
App. No.
17/208,707
Filed
Mar 22, 2021
Granted
Jun 30, 2026
Kind
B2
Art Unit
2125
USPC
706/15
Abstract

A method for monitoring operator compatibility within a deep learning framework, a computing system, and a non-transitory computer-readable storage medium are provided. The present disclosure relates to the field of deep learning. The method includes: generating first description information associated with at least one original operator and second description information associated with at least one modified operator, determining differences between the first description information associated with the at least one original operator and the second description information associated with the at least one modified operator, determining whether the differences satisfy a preset rule; and prompting information about incompatibility in response to determining that at least one of the differences does not satisfy the preset rule. The first and second description information are associated with the operator compatibility.

Claims (54)

1 . A method for monitoring operator compatibility within a deep learning framework, comprising:

accessing, by a processor in an electronic device, in a programming language environment and using a hash function, at least one original operator and its respective name, at least one modified operator and its respective name, first description information associated with the at least one original operator, and second description information associated with the at least one modified operator from a same storage space, wherein the first description information and the second description information are both associated with the operator compatibility;

determining, by the processor, differences between the first description information associated with the at least one original operator and the second description information associated with the at least one modified operator;

determining, by the processor, whether the differences satisfy a preset rule;

in response to determining that at least one of the differences does not satisfy the preset rule, automatically prompting, by the processor, information that the at least one modified operator is incompatible with the deep learning framework;

determining, by the processor, after prompting the information that the at least one modified operator is incompatible with the deep learning framework, whether to remodify the at least one modified operator associated with the at least one of the differences;

in response to determining to remodify the at least one modified operator associated with the at least one of the differences, generating, by the processor, a remodified operator and generating, by the processor, in the programming language environment and using the hash function, third description information associated with the remodified operator to be stored in the same storage space for being accessed using the hash function to determine whether the remodified operator is compatible with the deep learning framework; and

in response to determining not to remodify the at least one modified operator associated with the at least one of the differences, changing, by the processor, the information that the at least one modified operator is incompatible with the deep learning framework to that the at least one modified operator is compatible with the deep learning framework.

2 . The method according to claim 1 , wherein the preset rule comprises: a rule for determining whether the at least one modified operator is compatible within the deep learning framework.

3 . The method according to claim 1 , wherein in response to determining that at least one of the differences does not satisfy the preset rule, automatically prompting information that the at least one modified operator is incompatible with the deep learning framework comprises:

based on determining one of the differences that does not satisfy the preset rule, continuing to determine whether the remaining differences satisfy the preset rule, and prompting the information that the at least one modified operator is incompatible with the deep learning framework; or

based on determining one of the differences that does not satisfy the preset rule, forgoing determining whether the remaining differences satisfy the preset rule, and prompting the information that the at least one modified operator is incompatible with the deep learning framework.

4 . The method according to claim 1 , wherein the information that the at least one modified operator is incompatible with the deep learning framework comprises:

information about at least one of the differences that does not satisfy the preset rule; and

information about a modified operator associated with the at least one of the differences.

5 . The method according to claim 1 , wherein determining, after prompting the information that the at least one modified operator is incompatible with the deep learning framework, whether to remodify the at least one modified operator associated with the at least one of the differences comprises:

prompting, according to the information that the at least one modified operator is incompatible with the deep learning framework, to perform manual review to determine whether to remodify the at least one modified operator associated with the at least one of the differences.

6 . The method according to claim 1 , wherein the first and second description information comprise: an input, an output, and an attribute of an operator.

7 . The method according to claim 1 , wherein the first and second description information are stored in at least one of a dictionary, a table, or an array.

8 . A computing system, comprising:

a processor; and

a non-transitory memory that stores a program, the program comprising instructions that, when executed by the processor, cause the processor to:

access, in a programming language environment and using a hash function, at least one original operator and its respective name, at least one modified operator and its respective name, first description information associated with the at least one original operator, and second description information associated with the at least one modified operator from a same storage space, wherein the first description information and the second description information are both associated with the operator compatibility;

determine differences between the first description information associated with the at least one original operator and the second description information associated with the at least one modified operator;

determine whether the differences satisfy a preset rule;

in response to determining that at least one of the differences does not satisfy the preset rule, automatically prompt information that the at least one modified operator is incompatible with the deep learning framework;

determine, after prompting the information that the at least one modified operator is incompatible with the deep learning framework, whether to remodify the at least one modified operator associated with the at least one of the differences;

in response to determining to remodify the at least one modified operator associated with the at least one of the differences, generate a remodified operator and generate, in the programming language environment and using the hash function, third description information associated with the remodified operator to be stored in the same storage space for being accessed using the hash function to determine whether the remodified operator is compatible with the deep learning framework; and

in response to determining not to remodify the at least one modified operator associated with the at least one of the differences, change the information that the at least one modified operator is incompatible with the deep learning framework to that the at least one modified operator is compatible with the deep learning framework.

9 . The computing system according to claim 8 , wherein the preset rule comprises: a rule for determining whether the at least one modified operator is compatible within the deep learning framework.

10 . The computing system according to claim 8 , wherein in response to determining that at least one of the differences does not satisfy the preset rule, automatically prompting information that the at least one modified operator is incompatible with the deep learning framework comprises:

based on determining one of the differences that does not satisfy the preset rule, continuing to determine whether the remaining differences satisfy the preset rule, and prompting the information that the at least one modified operator is incompatible with the deep learning framework; or

based on determining one of the differences that does not satisfy the preset rule, forgoing determining whether the remaining differences satisfy the preset rule, and prompting the information that the at least one modified operator is incompatible with the deep learning framework.

11 . The computing system according to claim 8 , wherein the information that the at least one modified operator is incompatible with the deep learning framework comprises:

information about at least one of the differences that does not satisfy the preset rule; and

information about a modified operator associated with the at least one of the differences.

12 . The computing system according to claim 8 , wherein determining, after prompting the information that the at least one modified operator is incompatible with the deep learning framework, whether to remodify the at least one modified operator associated with the at least one of the differences comprises:

prompting, according to the information that the at least one modified operator is incompatible with the deep learning framework, to perform manual review to determine whether to remodify the at least one modified operator associated with the at least one of the differences.

13 . The computing system according to claim 8 , wherein the first and second description information comprise: an input, an output, and an attribute of an operator.

14 . The computing system according to claim 8 , wherein the first and second description information are stored in at least one of a dictionary, a table, or an array.

15 . A non-transitory computer-readable storage medium storing a program, the program comprising instructions that, when executed by a processor of a computing system, cause the computing system to:

access, in a programming language environment and using a hash function, at least one original operator and its respective name, at least one modified operator and its respective name, first description information associated with the at least one original operator, and second description information associated with the at least one modified operator from a same storage space, wherein the first description information and the second description information are both associated with the operator compatibility;

determine differences between the first description information associated with the at least one original operator and the second description information associated with the at least one modified operator;

determine whether the differences satisfy a preset rule;

in response to determining that at least one of the differences does not satisfy the preset rule, automatically prompt information that the at least one modified operator is incompatible with the deep learning framework;

determine, after prompting the information that the at least one modified operator is incompatible with the deep learning framework, whether to remodify the at least one modified operator associated with the at least one of the differences;

in response to determining to remodify the at least one modified operator associated with the at least one of the differences, generate a remodified operator and generate, in the programming language environment and using the hash function, third description information associated with the remodified operator to be stored in the same storage space for being accessed using the hash function to determine whether the remodified operator is compatible with the deep learning framework; and

in response to determining not to remodify the at least one modified operator associated with the at least one of the differences, change the information that the at least one modified operator is incompatible with the deep learning framework to that the at least one modified operator is compatible with the deep learning framework.

16 . The non-transitory computer-readable storage medium according to claim 15 , wherein in response to determining that at least one of the differences does not satisfy the preset rule, automatically prompting information that the at least one modified operator is incompatible with the deep learning framework comprises:

based on determining one of the differences that does not satisfy the preset rule, continuing to determine whether the remaining differences satisfy the preset rule, and prompting the information that the at least one modified operator is incompatible with the deep learning framework; or

based on determining one of the differences that does not satisfy the preset rule, forgoing determining whether the remaining differences satisfy the preset rule, and prompting the information that the at least one modified operator is incompatible with the deep learning framework.

17 . The non-transitory computer-readable storage medium according to claim 15 , wherein the information that the at least one modified operator is incompatible with the deep learning framework comprises:

information about at least one of the differences that does not satisfy the preset rule; and

information about a modified operator associated with the at least one of the differences.