Reputation: 137
Is it possible to share a model registry completely between Dev and Prod environment? So my idea is to create 10000 models in dev and maybe select 2000 from there to work in prod. I am planning to use AWS model registry. So if I do the training and testing and hyperparameter tuning in my AWS dev environment, is it possible to then share the registry in prod? The obvious reason is that it does not make sense to use the prod to do the training and testing again.
Please advise!
Thanks in advance!
Upvotes: 2
Views: 1639
Reputation: 2500
As of 2024 Amazon SageMaker Model Registry now supports cross-account machine learning (ML) model sharing and I do use this approach.
Old approach, I been using, was an export from one account model registry to another account model registry and overlapping artifacts all the time and made it more complicated for deployment on multi accounts and central models governance.
Now once you deploy sagemaker using for example terraform you can select cross account enablement for model registry and work with roles permissions only.
Here is detailed documentation with all needed parts to make it happen,
Topics:
Upvotes: 0
Reputation: 3355
I think AWS just released the cross account model registration feature not long ago, see: https://docs.aws.amazon.com/sagemaker/latest/dg/model-registry-version.html#model-registry-version-xaccount
Upvotes: 0
Reputation: 4037
It depends how you define Dev and Prod.
If by Dev and Prod you mean different AWS account (which is a good practice - see doc and blog), you cannot share fractions of a model registry from a given account to another account, but you can create triggers to export models from one model registry to another, as documented in this blog post https://aws.amazon.com/blogs/machine-learning/patterns-for-multi-account-hub-and-spoke-amazon-sagemaker-model-registry/
If your Dev and Prod live in the same AWS account and you are just looking for ways to differentiate them, you can use:
Upvotes: 2