How to create custom templates for model training and building in sagemaker studio?

0

I am going through documentation provided here , https://github.com/aws-samples/amazon-sagemaker-build-train-deploy and would like to create my own model building, training and deploying templates. can I download/clone the sagemaker provided templates and start modifying to my own needs. I understand , we need to set up a aws catalog portfolio and products under it, to be able to use such templates. my question is which project do i need to clone and modify , say if i want to build my own training and model building template, which particular code base or code file do i need to change. I assume , the train.py file here -> https://github.com/aws-samples/amazon-sagemaker-build-train-deploy/blob/master/08_projects/modelbuild/pipelines/endtoendmlsm/train/train.py would be one i will customize. but once i change these files, how do i use them. how would i set up my custom template, create a product under my catalog? if yes, how to I link my custom train.py code to this new custom sagemaker project template. the documentation or samples only show how to use the prebuild templates under sagemaker project templates, but how do i get my own template pushed there ? can this be done via say cloudformation ?

gefragt vor 2 Jahren426 Aufrufe
1 Antwort
0

Hi,

This workshop works through the typical ML process to build and deploy a model to predict the fault on machine based on synthetic datasets. To start building it, you will need to have a AWS account, the link is here.

Once you have an account, you can start from module 1 from this step. I would recommend follow each module in order and try to implement in your account. This will help you understand the methodology and familiar with the service.

To modify it, you will need to collect and use your datasets, identifying your business problem. You could try to re-use the same algorithm in the module but it depends on the issue you try to resolve.

Hope it helps

Thanks,

AWS
Jady
beantwortet vor 2 Jahren
profile picture
EXPERTE
überprüft vor einem Monat

Du bist nicht angemeldet. Anmelden um eine Antwort zu veröffentlichen.

Eine gute Antwort beantwortet die Frage klar, gibt konstruktives Feedback und fördert die berufliche Weiterentwicklung des Fragenstellers.

Richtlinien für die Beantwortung von Fragen