CODEWITHSUNDEEP
amazon-web-servicesjupyter-notebookinstanceamazon-sagemaker
Borut Flis
Borut Flis

Reputation: 16415

What sagemaker algorithm can be trained by the minimal instance?

I am following this example on how to train a machine learning model in Amazon-sagemaker. The problem is ml.t2.medium instance fail to satisfy the constraint of K-means algorithm. I can't use the instances that satisfy the constraints of the algorithm(I will ask support for allocation of this resource). My question is there any algorithm from sagemaker that one can train on the MNIST dataset with the minimal instance?

Upvotes: 0

Views: 464

Answers (1)

Chris Williams
Chris Williams

Reputation: 35258

Edit

Noticed when browsing pricing that for training the minimum size is ml.m5.large so you will need to look through these to determine which instance type to use.

AWS do provide in their free tier "50 hours of m4.xlarge or m5.xlarge for training".

Original

This comes down to a number of factors.

I am not aware of any specific limitations based on algorithms in fact AWS recommend general instance families on this page, with the k-means algorithm this is targeted at CPU or GPU based instances (so it does not exclusively require GPU).

There are a list of available instance types in the documentation that do not mention the instance type you're referring to, I would recommend trying to use the ml.t3.medium instance type instead. This is a newer generation and likely to be supported more across current version of the algorithms.

Upvotes: 1

Related Questions