IP Library Patent Application 18397909
Patent Application
App. No. 18/397,909

SYSTEMS AND METHODS FOR AN ACCELERATED AND ENHANCED TUNING OF A MODEL BASED ON PRIOR MODEL TUNING DATA

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 None
App. No.
18/397,909
Abstract

Disclosed examples including generating a joint model based on first and second subject models, the first and second subject models selected based on a relationship between the first and second subject models; selecting the joint model from a plurality of joint models after a determination that entropy data points of the joint model satisfy a threshold, the entropy data points based on multiple tuning trials of the joint model; and providing tuning data associated with the joint model to a tuning session of a target model.

Claims (56)

1 . An apparatus comprising:

interface circuitry;

instructions, and

programmable circuitry to be programmed by the instructions to:

generate a joint model based on first and second subject models, the first and second subject models selected based on a relationship between the first and second subject models;

select the joint model from a plurality of joint models after a determination that entropy data points of the joint model satisfy a threshold, the entropy data points based on multiple tuning trials of the joint model, and

provide tuning data associated with the joint model to a tuning session of a target model.

2 . The apparatus of claim 1 , wherein the programmable circuitry is to:

access the first and second subject models in a pool of historical subject models; and

select the first and second subject models as a pair of subject models to generate the joint model after a relatedness metric value corresponding to a relatedness between the first and second subject models satisfies a relatedness threshold.

3 . The apparatus of claim 2 , wherein the programmable circuitry is to:

determine a second relatedness metric value for a third subject model and a fourth subject model from the pool of historical subject models; and

after a determination that the second relatedness metric value does not satisfy the relatedness threshold, not select the third and fourth subject models as a pair of subject models to generate the joint model.

4 . The apparatus of claim 1 , wherein the programmable circuitry is to:

compare a behavior of the first subject model to a behavior of the second subject model; and

select the first and second subject models based on the comparison.

5 . The apparatus of claim 1 , wherein the tuning data associated with the joint model is historical tuning data associated with the first and second subject models.

6 . The apparatus of claim 1 , wherein the programmable circuitry is to:

generate predictions based on the multiple tuning trials of the joint model; and

generate the entropy data points based on at least some of the predictions.

7 . The apparatus of claim 1 , wherein the entropy data points of the joint model are maximum entropy data points.

8 . A storage device or storage disk comprising computer-readable instructions to cause programmable circuitry to at least:

generate a joint model based on first and second subject models, the first and second subject models selected based on a relationship between the first and second subject models;

select the joint model from a plurality of joint models after a determination that entropy data points of the joint model satisfy a threshold, the entropy data points based on multiple tuning trials of the joint model; and

provide tuning data associated with the joint model to a tuning session of a target model.

9 . The storage device or storage disk of claim 8 , wherein the computer-readable instructions are to cause the programmable circuitry to:

access the first and second subject models in a pool of historical subject models; and

select the first and second subject models as a pair of subject models to generate the joint model after a relatedness metric value corresponding to a relatedness between the first and second subject models satisfies a relatedness threshold.

10 . The storage device or storage disk of claim 9 , wherein the computer-readable instructions are to cause the programmable circuitry to

determine a second relatedness metric value for a third subject model and a fourth subject model from the pool of historical subject models; and

after a determination that the second relatedness metric value does not satisfy the relatedness threshold, not select the third and fourth subject models as a pair of subject models to generate the joint model.

11 . The storage device or storage disk of claim 8 , wherein the computer-readable instructions are to cause the programmable circuitry to:

compare a behavior of the first subject model to a behavior of the second subject model; and

select the first and second subject models based on the comparison.

12 . The storage device or storage disk of claim 8 , wherein the tuning data associated with the joint model is historical tuning data associated with the first and second subject models.

13 . The storage device or storage disk of claim 8 , wherein the computer-readable instructions are to cause the programmable circuitry to:

generate predictions based on the multiple tuning trials of the joint model; and

generate the entropy data points based on at least some of the predictions.

14 . The storage device or storage disk of claim 8 , wherein the entropy data points of the joint model are maximum entropy data points.

15 . A method comprising:

generating a joint model based on first and second subject models, the first and second subject models selected based on a relationship between the first and second subject models,

selecting, by executing an instruction with programmable circuitry, the joint model from a plurality of joint models after a determination that entropy data points of the joint model satisfy a threshold, the entropy data points based on multiple tuning trials of the joint model, and

providing tuning data associated with the joint model to a tuning session of a target model.

16 . The method of claim 15 , including:

accessing the first and second subject models in a pool of historical subject models; and

selecting the first and second subject models as a pair of subject models to generate the joint model after a relatedness metric value corresponding to a relatedness between the first and second subject models satisfies a relatedness threshold.

17 . The method of claim 16 , including:

determining a second relatedness metric value for a third subject model and a fourth subject model from the pool of historical subject models, and

after a determination that the second relatedness metric value does not satisfy the relatedness threshold, not selecting the third and fourth subject models as a pair of subject models to generate the joint model.

18 . The method of claim 15 , including:

comparing a behavior of the first subject model to a behavior of the second subject model, and

selecting the first and second subject models based on the comparison.

19 . The method of claim 15 , wherein the tuning data associated with the joint model is historical tuning data associated with the first and second subject models.

20 . The method of claim 15 , including:

generating predictions based on the multiple tuning trials of the joint model; and

generating the entropy data points based on at least some of the predictions.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 11, 2026
From: INTEL CORPORATION
To: INTEL PRODUCTS IP LLC
Reel/Frame 075991/0981 →