Using a client to manage remote machine learning jobs

Patent metadata

Field	Value
Publication number	US-10621019-B1
Application number	US-201815919179-A
Country	US
Kind code	B1
Filing date	Mar 12, 2018
Priority date	Nov 22, 2017
Publication date	Apr 14, 2020
Grant date	Apr 14, 2020

Abstract

Methods, apparatuses, and systems for a web services provider to interact with a client on remote job execution. For example, a web services provider may receive a job command, from an interactive programming environment of a client, applicable to job for a machine learning algorithm on a web services provider system, process the job command using at least one of a training instance and an inference instance, and provide metrics and log data during the processing of the job to the interactive programming environment.

First claim

What is claimed is:

1. A computer-implemented method comprising:
receiving a start job command, from an interactive programming environment, to start a training job for a machine learning algorithm on a web services provider system, wherein the start job command is generated by the interactive programming environment and includes an indication of a location of a container storing the machine learning algorithm, a location of training data, a location to store a model, resources to be used by the container, and a stopping condition;
accessing the training data to be used to train the machine learning algorithm;
instantiating the container storing the machine learning algorithm;
running the job using the instantiated container to train the machine learning algorithm using the accessed training data and produce the model; and
providing metrics and log data during the running of the job to the interactive programming environment.

2. The computer-implemented method of claim 1 , wherein the start job command further includes a job name.

3. A computer-implemented method comprising:
receiving a job command, from a client, applicable to a job for a machine learning algorithm on a web services provider system, wherein the job command includes an indication of a location of a container storing the machine learning algorithm, a location of training data, a location to store a model, resources to be used by the container, and a stopping condition;
processing the job command using at least one of a training instance and an inference instance, wherein processing the job command comprises:
accessing the training data to be used to train the machine learning algorithm, and instantiating the container storing the machine learning algorithm; and
providing metrics and log data during the processing of the job to the client.

4. The computer-implemented method of claim 3 , wherein the job command is a command to start the job for the machine learning algorithm and further includes a job name.

5. The computer-implemented method of claim 4 , wherein processing the job command further comprises training the machine learning algorithm using the training data in the training instance.

6. The computer-implemented method of claim 4 , wherein processing the job command further comprises training the machine learning algorithm using feature processed training data in the training instance.

7. The computer-implemented method of claim 4 , wherein processing the job command further comprises evaluating the machine learning model using evaluation data in the inference instance.

8. The computer-implemented method of claim 3 , wherein the job is a training job for the machine learning model and the job command is a command to stop the training job for the machine learning algorithm and includes a job name.

9. The computer-implemented method of claim 3 , wherein the job command is a command to stop an evaluation job for the machine learning model and includes a job name.

10. The computer-implemented method of claim 9 , wherein processing the job command further comprises training the machine learning algorithm using training data in the training instance.

11. The computer-implemented method of claim 9 , wherein processing the job command further comprises evaluating the machine learning model using evaluation data in the inference instance.

12. The computer-implemented method of claim 3 , wherein the job command is a command to modify an existing job for the machine learning algorithm and further includes a job name, hyperparameters, and resources to be used, wherein at least one of the locations, hyperparameters, and resources has changed from an initial indication.

13. A system comprising:
a first one or more physical computing devices of a service provider network to execute training instances that train machine learning (ML) models and to execute inference instances that host ML models;
a second one or more physical computing devices to implement a control plane of the service provider network, the control plane comprising instructions which when executed cause the control plane to:
receive a job command, from a client device, applicable to a job for a machine learning algorithm, wherein the job command is generated by an interactive programming environment executed by the client device and includes an indication of a location of a container storing the machine learning algorithm, a location of training data, a location to store a model, resources to be used by the container, and a stopping condition;
cause the job command to be processed using at least one of a training instance or an inference instance, wherein causing the job command to be processed comprises:
accessing the training data to be used to train the machine learning algorithm, and
instantiating the container storing the machine learning algorithm; and
provide metrics and log data during the processing of the job to the programming environment.

14. The system of claim 13 , wherein the job command is a command to start the job for the machine learning algorithm and further includes a job name.

15. The system of claim 14 , wherein to cause the job command to be processed further comprises to train the machine learning algorithm using training data in the training instance.

16. The system of claim 14 , wherein to cause the job command to be processed further comprises to train the machine learning algorithm using feature processed training data in the training instance.

17. The system of claim 14 , wherein to cause the job command to be processed further comprises to evaluate the machine learning model using evaluation data in the inference instance.

18. The system of claim 14 , wherein the job is a job training job for the machine learning model and the job command is a command to stop the training job for the machine learning algorithm and includes a job name.

19. The system of claim 13 , wherein the job is the evaluation job for the machine learning model and job command is a command to stop the evaluation job for the machine learning model and includes a job name.

Assignees

• Amazon Tech Inc

Inventors

• Faulhaber Jr Thomas Albert
• Dirac Leo Parker

Classifications

• G06F9/455 Primary

  Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines · CPC title

• G06F2009/45575

  Starting, stopping, suspending or resuming virtual machine instances · CPC title

• G06F9/547 Primary

  Remote procedure calls [RPC]; Web services · CPC title

• G06F9/45558

  Hypervisor-specific management and integration aspects · CPC title

• G06N20/00

  Machine learning · CPC title