IP Library › Granted Patent US 11,361,253
Granted Patent B2
US 11,361,253 · App. 17/405,888 · Granted Jun 14, 2022

Modularized model interaction system and method

Inventors: Williams Falcon (New York, NY); Adrian Wälchli (Bern, CH); Thomas Henri Marceau Chaton (London, GB); Sean Presley Narenthiran (London, GB)
Assignee: Grid.ai, Inc.
G06N20/00G06F9/5005
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 11,361,253
App. No.
17/405,888
Granted
Jun 14, 2022
Kind
B2
Abstract

A modularized model interaction system and method of use, including an orchestrator, a set of hardware modules each including a standard set of hardware submodules with hardware-specific logic, and a set of model modules each including a standard set of model submodules with model-specific logic. In operation, the orchestrator determines a standard set of submodule calls to the standard submodules of a given hardware module and model module to implement model interaction on hardware associated with the hardware module.

Claims (24)

1. A system for modular machine learning model interaction, comprising:

a processing system; and

a non-transitory computer readable medium storing instructions executable by the processing system, comprising:

a set of model modules, each comprising model-specific logic for each of a set of standard model submodules;

a set of computing resource modules, each associated with a different computing resource and comprising computing resource-specific logic for each of a set of standard computing resource submodules; and

an orchestrator configured to:

a) initialize a computing resource by executing a series of standard computing resource submodules from an associated computing resource module; and

b) initialize a model on the computing resource by executing a series of standard model submodules from an associated model module.

2. The system of claim 1 , wherein the orchestrator is configured to perform a) and b) for different model and computing resource combinations, using:

different model modules associated with the respective model of each combination;

different computing resource modules associated with the respective computing resource of each combination;

the same series of standard model submodules; and

the same series of standard computing resource submodules.

3. The system of claim 1 , wherein interacting with the model comprises at least one of: training the model, validating the model, testing the model, or predicting using the model.

4. The system of claim 3 , wherein the standard model submodules comprise a train submodule, a validate submodule, and a test submodule, wherein each model module comprises model-specific training logic associated with the train submodule, model-specific validation logic associated with the validate submodule, and model-specific testing logic associated with the test submodule.

5. The system of claim 4 , wherein the test submodule is executed using “training_step( )”, wherein the validate submodule is executed using “validation_step ( )”, and wherein the test submodule is executed using “test_step( )”.

6. The system of claim 1 , wherein the orchestrator comprises a set of model interaction modules, each configured to execute an interaction-specific series of standard model submodules.

7. The system of claim 6 , wherein a model interaction module of the plurality comprises a standardized loop, wherein the same standardized loop is executed to interact with different models.

8. The system of claim 7 , wherein the standard model submodules of each model module comprise model- and interaction-specific logic for a single loop iteration.

9. The system of claim 1 , wherein the computing resource-specific logic comprises driver-specific calls to hardware drivers for the respective computing resource.

10. The system of claim 1 , wherein the computing resources comprise at least one of: a central processing unit (CPU), a graphics processing unit (GPU), an image processing unit (IPU), or a tensor processing unit (TPU).

11. The system of claim 1 , further comprising a set of data precision modules, each comprising precision-specific logic, wherein the orchestrator is configured to initialize the computing resource using the precision-specific logic of a data precision module specified by a user.

12. The system of claim 1 , further comprising a set of distribution modules, each comprising inter-computing-resource coordination logic, wherein the orchestrator is configured to initialize a set of computing resources and orchestrate model interaction across the set of computing resources using the inter-computing-resource coordination logic for a scaling module specified by a user.

13. The system of claim 1 , wherein the computing resource modules, the model modules, and the orchestrator are separate and distinct.

Assignments (2)
SECURITY INTEREST Recorded Feb 3, 2026
From: GRID.AI, INC.
To: JPMORGAN CHASE BANK, N.A.
Reel/Frame 073679/0233 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 3, 2021
From: FALCON, WILLIAMS; WÄLCHLI, ADRIAN; CHATON, THOMAS HENRI MARCEAU; NARENTHIRAN, SEAN PRESLEY
To: GRID.AI, INC.
Reel/Frame 057380/0566 →
Continuity (4)
Provisional Application 63067203 · Aug 18, 2020
Provisional Application 63148819 · Feb 12, 2021
Provisional Application 63166629 · Mar 26, 2021
Related Publication 20220058523A1 · Feb 24, 2022
Cited By (2)
US 12,254,295 US 12,530,145